Clean up language strings, add openomf-translations submodule

[Nopey]

this PR:

• Generates an unnamed enum full of lang string indexes like LangPilot
• Replaces all uses of lang_get + hardcoded value with a variant from said enum.
• Reduces translator burden by not shipping unused language strings.
• Ships a single localization file per language (not counting tournament localization, that's another PR's problem)
• Reduces runtime memory usage by retaining the useful language strings' data in a single contigous allocation, and discarding their description.

Rationale for using cmake codegen instead of a straight C enum:

• Since we're no longer generating ENGLISH.LNG from OMF ENGLISH.DAT, this is no longer relevant-- it is now an enum with two C macros in languages.h!

OMF ENGLISH.DAT consumes 65,397 bytes of memory after discarding the descriptions
A cut down OpenOMF ENGLISH.LNG consumes 35,257 bytes of memory and only contains 458 strings.

[katajakasa]

This seems messy and difficult. Would it be better to just create our own language files, add only text that is used and include them to the project? Then we could drop the original language files completely.

[Nopey]

Added "Add language string for Text Encoding", so there's something signalling whether the .LNG is in utf-8 or cp437 EDIT: removed that commit
Rebased on master to resolve merge conflict with stdbool-everywhere pr.

Intentionally not removing the ENGLISH.DAT + scripts we wrote -> ENGLISH.LNG pipeline, because despite its complexity it allows us to avoid commiting OMF assets into the repo. EDIT: those assets now live in the openomf-translations repo

Should I remove the Dansk Maskinoversottelse? its derivative of ENGLISH.DAT. EDIT: removed.

[Nopey]

the resources/translations submodule is checked out to omf2097/openomf-translations#1

[mrannanj]

Still undecided about submodules, but initially okay.