When uncompressed it requires less bandwidth and it is more suitable for
compression and/or further processing.

+ fail if audio BPS or sample rate is invalid rather than displaying
  only a warning
+ removed unused FRAME_TIMEOUT

Incorrect BPP (3) was used for RGBA

- use $(CXX) instead of hard-codec g++ in Makefile.convert
- conversion -> list-conversions + brief usage

fix type o found to find

+ small improvements (use of MAX_PATH_SIZE)

SDL, GL and NDI required mainloop only on macOS.

The export structure was not zero-initialized as expected when checked
for NULL.

Use an user-friendlier error message than assert.

Removed msg_change_fec_data, use msg_universal instead.

This is a more versatile solution allowing further extensions without
a need to modify the global structure at an expese of worse type control.
As a replacement, a tag starting msg_universal::text is used/suggested.

1. set the volume for all channels
2. correctly set attribute muted according to the value (until now it
   was actually never set)

- show actual playback volume as 0 (-inf) if muted receiver
- display explicitly receiver mute status
- "Received audio volume" -> "Playback volume" - this has more clear
  semantics and is consistent with message written when '*'/'/'/'m' is
  pressed.

Do not show hint to use NAT traversal if user uses private or loopback
address as a peer address.

This avoids misleading displaying in otherwise legitimate cases -
transmissions over loopback or behind NAT but just inside LAN.

Use UINT32_MAX for sizes first. This may help if WAV file is incorrectly
closed indicating that the size values aren't valid (at least FFmpeg
understands it so).

+ fixed message for overflow

This modernizes the code and makes the semantic of prototypes more clear.

Enable querying location of executable

-add function get_exec_path and C++ wrapper around it

Do not assume that initialization of null device must succeed - eg. in
some errorneous setup there may be API mismatch leading to init failed.

Was printed NULL ("(null)") because *fmt was set to NULL early.

[rtp] fix display_buf_increase_warning warning

rtp_set_recv_buf returns TRUE on success, which is #define'd to 1. Fix the logic in the check so the warning does not print erroneously.