Fix build on WSL and remove Python

[jmeggitt]

The previous version would attempt to fetch the required headers from every include folder, overwriting the previous results as it went and in my case, stumbling over broken symlinks in duplicate folders.

The python script would return all of the include folders (example result):

# Include directories given by python script (example from when I ran it): 
*/arm-frc2019-linux-gnueabi/usr/lib/gcc/arm-frc2019-linux-gnueabi/6.3.0/include
*/arm-frc2019-linux-gnueabi/usr/lib/gcc/arm-frc2019-linux-gnueabi/6.3.0/include-fixed
*/arm-frc2019-linux-gnueabi/usr/include

# The includes we actually care about:
*/arm-frc2019-linux-gnueabi/usr/include
*/arm-frc2019-linux-gnueabi/usr/lib/**/include/stddef.h

Since the python script just ran a command and trimmed the results, I moved the functionality to the makefile instead and was able to make it more verbose as a result.

[auscompgeek]

Is this correct, or should be $(CXX_FRC)?

[jmeggitt]

Sadly, thats correct. Makefiles have weird syntax and work with their own variable system.

If I were to try referencing an environment variable the regular way, it would start parsing and give up when it realizes there are no makefile variables that could fit. So if I had defined CXX2 before hand, it would give up half way through and delete the characters up to the point where it failed. The resulting line would become toolchain := _FRC. A second $ acts as an escape character and allows it to be read like any other shell command. Thats also why I included two on line 20.

[auscompgeek]

Oh, right. I think the better way to deal with this would be to use ?=.

[auscompgeek]

Say, why do we even copy headers in the first place? The correct way to get clang to target non-host is to pass -target.