-here a scaling factor is sampled in addition to the uncertainty.
The additional difference with BayesSP is that the experimental
data are here given in input, while in BayesSP the input is already
the squared difference.

-some code optimisation

-regtest for both

allows syntax to be checked

Includes both syntax and autocompletion

This is actually a squash of many commits entirely made
by Gareth.

It adds the possibility to generate a file useful for VIM syntax highlighting.

[makedoc]

this is usually not important, but for the case (like with
metainference) in which you have a bias that is applied on
a large number of arguments, then it was getting very slow

- can now work also with only 1 replica
- made regtest to cover more code
- fix a bug in stat initialisation when no parameters where provided

[makedoc]

See #186