Module interaction and state fixes

[Gumball2415]

This pull request aims to fix module state and interaction bugs.

This pull request iterates upon PR #195.

Changes in this PR:

• Avoid checking pattern editor assert with unsigned integer cast
  • Fixes Assertion fail causing crash #209.
• Wait for APU mutex lock during .wav export
  • Fixes .wav export skips ticks #206.
• Assert legacy mixing levels and ranges
  • Fixes Are channels other than apu1 and apu2 being quiet in multichip modules intended? #213.
• Fix detune offset direction
  • Fixes N163 pitch LUT not detuned properly with custom detuning table #225.
• Fix incorrect speed in PAL NSF exports
  • Fixes Incorrect PAL speed in NSF exports #223.
• Fix access violation in MRU submenu list update
  • Fixes x64-build crash with NoRecentDocsHistory enabled in registry #243.
• Avoid division by zero in MML sequence parsing
  • Fixes Zero-width volume MML macro crashes the program #222.
• Fix effect number input using numpad
  • Fixes [Feature Request] Restore ability to input effects using number pad #48.
• Disable Custom Exporter DLL loading
  • Fixes Possible Custom Exporter DLL exploit vulnerability #232.
• Force modules to be saved as Dn-FT modules
  • This prevents modules to be saved with version 0x450 but without the Dn-FT specific file header, causing the program to interpret it as FT 050b modules instead.
  • Fixes complications from fixing Specific effects changing type when older modules are saved in 0.5.0.0 #184.

[nyanpasu64]

this can't be right? i saw a frame -1 causing a crash at runtime.

i have determined that famitracker has crashed trying to restore channel state as of frame (unsigned)-1 = 0xffffffff, on channel 4 (presumably it would crash on all channels and 4 is the first one restored)

famitracker wine crash.pdf

try a release ASAN build with the assert removed, if it crashes, then this PR is wrong.

[Gumball2415]

it does not crash on ASAN release with the asserts removed