Improve lmms::fastRand() and use it instead of std::rand()

[rubiefawn]

• Combine overloads of lmms::fastRand() into one template
  • Remove RAND_MAX constant since it is not used outside lmms::fastRand() (and all places that did previously use it should have been calling lmms::fastRand() anyway)
  • This is also done to lmms::fastLog10f() for consistency, though it is not related to the goal of the PR
• Add lmms::oneIn() convenience function to replace occurences of (fastRand() % chance == 0)
• Replace uses of std::rand() with lmms::fastRand() for consistency
• Replace uses of scale * lmms::fastRand() / RAND_MAX - offset with lmms::fastRand(min, max)

// Old
int x = static_cast<int>(10.0f * rand() / RAND_MAX - 3.0f);

// New
int x = fastRand(-3, 7);

Testing

• Triple Oscillator noise sounds the same
• Organic randomize preset button works
• Multiple Peak Controllers can coexist
• Sfxr randomize preset button works
• SID plugin sounds the same
• Vibed plugin sounds the same
• .ds samples with noise sound the same
• Undo/redo doesn't instantly explode
• Mixer channel color randomization works
• Clip color randomization works
• Track color randomization works

[messmerd]

I haven't reviewed everything yet, but here's what I have so far

[bratpeki]

Briefly tested Mallets, Spectrum Analyzer, ReverbSC, Flanger and Ogranic. I think we can (finally) merge! 🎉

[szeli1]

I'm changing my approval in light of std::nearbyint.

Dismissing by request on Discord

[regulus79]

I was asked to review this PR; I have looked through all the code, and the changes appear fine to me. I am not an expert in pseudorandom number generation, but assuming people have tested it, I think it's fine.

[rubiefawn]

Now that this has an approval from a dev with write access, I plan to merge this PR on 2026-01-08, provided no issues are found in the meantime. If anyone else would like to review this PR, please do so before that date.

[messmerd]

Quick review - haven't looked at things too closely yet