gmewav: convert NSF, GBS, SPC, IT, XM, S3M, MOD to WAV

$ ./gmewav --help
usage: gmewav [options] vgmfile wavfile

options:
  -h, -?, --help        show this usage info
  -t LENGTH             render LENGTH seconds of audio (default: 150.0)
  -f LENGTH             fade out the last LENGTH seconds (default: 0.0)
  -m MOVEMENT           render movement MOVEMENT (default: 1)

Supported game music formats (Game_Music_Emu):
vgm, gym, spc, sap, nsf, nsfe, ay, gbs, hes, kss
Supported module formats (Dynamic Universal Music Bibliotheque):
it, xm, s3m, mod

// Issue "Mute": --mute and --solo to extract individual tracks
// Voice numbers: 0 <= i < count
// Mute state: 1 for silent, 0 for playing
// Muting masks: (1 << i) OR'd together, where 0x04 mutes voice 2
// Because
int gme_voice_count(const Music_Emu *self);
const char *gme_voice_name(const Music_Emu *self, int i);
const char *gme_mute_voice(Music_Emu *self, int i, int mute);
const char *gme_mute_voices(Music_Emu *self, int muting_mask);

// DUMB appears to lack an API to query a DUH for voice count.
// The duh_sigrenderer_get_n_channels() call returns how many
// output channels it is generating (2 for stereo), not how many
// voices are used.  Solo should treat all DUHs as having
// DUMB_IT_N_CHANNELS channels.
void dumb_it_sr_set_channel_muted(
  DUMB_IT_SIGRENDERER *self, int channel, int muted
);

// gme_identify_file() places non-NULL in out iff the file is
// identifiable.
// For listing GME formats, I must file an upstream issue to add a
// C API call to get the canonical extension (not the console name)
// of a gme_type_t.
// The version of DUMB in Ubuntu 16.04 LTS provides no anaogous
// extensibility means; dumb_load_any() isn't in LTS yet.
gme_type_t const* gme_type_list();
int gme_type_multitrack(gme_type_t self);  // nonzero if multiple movements
const char *gme_identify_file(const char *path, gme_type_t *out);
const char* gme_type_system(gme_type_t self);

// TODO: Figure out which RIFF chunks to use as tags
struct gme_info_t {
  // Contains these members and others:
  const char *system, *game, *song, *author, *copyright, *comment, *dumper;
};

// Allocates a gme_info_t containing tags as C strings
const char *gme_track_info(const Music_Emu *self, gme_info_t** out, int track );
void gme_free_info(gme_info_t *self);

// Set to DUMB_RQ_ALIASING, DUMB_RQ_LINEAR, DUMB_RQ_CUBIC
// I plan to have DUMB's "aliasing" mode render at double sample rate
// (that is, 88200 Hz), convolve audio with the low-pass kernel
// [-1 0 9 16 9 0 -1]/32, and decimate by 2.  This way a chiptune still
// sounds crisp, but without folding quite as many overtones over fs/2.
int dumb_resampling_quality = DUMB_RQ_LINEAR;

./gmewav -m 2 -S 1 -t 150 something.nsf movement2-pulse1.wav
./gmewav -m 2 -S 2 -t 150 something.nsf movement2-pulse2.wav
./gmewav -m 2 -S 3 -t 150 something.nsf movement2-triangle.wav
./gmewav -m 2 -S 4 -t 150 something.nsf movement2-noise.wav
./gmewav -m 2 -S 5 -t 150 something.nsf movement2-dmc.wav