`except:` might catch the wrong exception by mistake

In addition to numpy arrays, I make it possible to pass plain python arrays.

I think the inclusion of numpy headers was related to non used variables
forgotten from an intermediate implementation.
I just removed them. In this manner it is not necessary that numpy is
already installed when plumed extension is compiled.

Numpy would be anyway required to run our test.

Sets the deployment target to 10.9 when compiling on version 10.9 and above.
The fix is taken from https://github.com/pandas-dev/pandas/pull/24274/files

GB: removed whitespace change and merged. Closes #445

Thanks @valsson

I removed the newline at the end of the string

there should be a function called test(), otherwise nose just imports
the modules and thinks no test was done

Pypi ships also the plumed.cpp file, so it is not necessary to
ask users to install cython

I added some material needed for pypi packaging.

I made a few more changes to setup.py in order to work with MacPorts.
This allowed me to add a temporary Portfile for py-plumed
that is now used in travis tests.

The plan is to submit this Portfile to MacPorts after we release v2.5

Notice that one test is duplicated (python/test) in order to be available
when using nosetests from python.

- renamed script to setup.py
- Use "make pip" to copy required files in this dir for pip
- Use try/except when importing numpy and cython so as to notify the user
- renamed env vars to plumed_xxx so that there is no risk of clashes

Plumed can now also be installed using

cd python
make pip
pip install --user .

make pip is just needed to bring version information and Plumed.h
in this directory. Might be solved differently in MacPorts / Pypi.

This allows creating plumed objects with

p=plumed.Plumed(kernel="/path/to/libplumedKernel.so")

Not strictly necessary, but might be useful to allow more
flexibility.

It does not link to library anymore, but just includes Plumed.h file.
It thus could be compiled with different compilers and flags wrt plumed.

Notice that as of PLUMED 2.5 Plumed.h also contains the full implementation
of the dynamic loader, so it is sufficient to define the proper cpp
flags in order to recompile the loader and avoiding the need to link
plumedWrapper.a.

This basically allows the python wrapper to be selfcontained and installable
before having installed plumed.

I think python should use the C++ interface rather than the C one.
The reason is that the C++ interface included in PLUMED 2.5
rethrows exceptions and thus can be used to load
a shared library compiled with a different compiler or
linked against a different standard library.

In perspective, this will allow compiling PLUMED with
a different compiler/library than that used to compile
python itself.

Set LD_LIBRARY_PATH before running plumed when compiling python
interface

Now that libplumedKerneldylib does not depend anymore on libplumed.dylib
it is possible to use two-level namespace on osx.

This change has two implications:
- the shared library used to resolve symbols during linking is recorded
  and used at runtime.
- all symbols should be available at link time

A side effect of this is that while compiling dylibs for the LOAD
action we should link them to the libplumedKernel as well.
For consistency I do the same on Linux, though it is not necessary in principle.

This should ensure that plumed_finalize is called when the python/cython
class is destroyed

Addresses #295

Variables are passed as environment variables from Makefile,
where they can be constructed using Makefile variables.

This solves some problems I had with python2.7 not able to parse
correctly the Makefile.conf file. In addition, it looks like correct
directories are used also when plumed is already installed.

Also notice that the plumed executable is now only invoked
to retrieve the version number and that it is invoked with the
correct path, so that a previously installed plumed will
not be in the way.

It was giving troubles on my mac with py2.7

Use ./configure PYTHON_BIN=python2.7 to specify a version. Can
be either relative path or full path. If undefined, configure will
search from a list of names (currently, only name "python" is searched).

Notice that cython command is not searched, since it is not actually used
to build the interface. The test is rather done on the availability
of the Cython module and other required modules.

Moreover, I moved the test down in the list, after all the libraries

* Added first working python interface

* Tidied up python interface

* Further simplification of python

* Added example which should be deleted later

* Added functions for getting values from plumed to python

* Added regression test for python interface

* Added mechanism to return exceptions from plumed to python

* First commit of experimental stuff to catch exceptions in python correctly

* Got rid of unecessarily duplicated method in PlumedMain

* Fixed crazy MAC input encoding

* Oops, had also broken the script..

* Removed debug messages. Works on Ubuntu Linux

Linus rules. Apple sucks.

* Fixed PlumedMain passing of data from plumed to outside code

* Hard coded real precision for python interface - this should have been with the last commit

* Added first go at working makefile for python interface

* Added options to configure to build python interface from makefile

* Small changes to where loading of python interface is done

* Removed unecessary wrapper c++ function from python interface

* Made python install work in a way that is consistent with the rest of plumed

* Added pythonpath to travis.yml

* Added installation of Cython to travis.yml

* Added ifdef block to stop cppcheck throwing an error

* Using uniqu_ptr for DataFetching object in PlumedMain

* Added install command for numpy to travis

* Added string to tell me whether plumed+python is being built

* Small change in travis.yml file to see if I can get python compile working

* Compilers for cython are now specified from Makefile.conf

* Now printing stuff on environment to try to work out why linking doesn't work

* A new attempt to fix the travis-ci issues

* Got rid of troubling TRAVIS_PYTHON_FLAG

* Small change to see if we can get this to work on travis

* Now setting LDSHARED from LDSO in makefile

* Added something to print everything in the python path when we run the python test

* Fixed python path in travis.yml

* Removed stuff for debugging configuration on travis

* Customized conda install (for python) in travis.yml so that osx version is installed for osx and linux version is installed for linux

* Set TMPDIR in .travis.yml to hopefully get osx to compile PLUMED + Python

* Made it so that python interface is rebuilt when user does make install so that correct paths are used

* Made it so that python interface is rebuilt in staging area during install procedure

* Corrected regtest script so that python tests are dealt with the same as everything else