fix AY freq-to-tone calculation

[ahihi]

fixes #348, and by extension https://codeberg.org/musikinformatik/SuperDirt/issues/281

AY.freqtotone is implemented with an "approximate empirical" formula:

sc3-plugins/source/AYUGens/sc/AY.sc

Lines 19 to 23 in 1c9bb52

	*freqtotone { |freq|
	// Approximate empirical...
	//^(109300 / (freq - 3.70727))
	^(110300 / (freq - 0.5))
	}

it is reasonably accurate at a sample rate of 44100 Hz, but at other rates it is severely out of tune.

i noticed that the ayemu code calculates a value ChipTacts_per_outcount, which depends on the chip frequency of the emulator as well as the sample rate:

sc3-plugins/source/AYUGens/AY_libayemu/src/ay8912.c

Line 369 in 1c9bb52

ay->ChipTacts_per_outcount = ay->ChipFreq / ay->sndfmt.freq / 8;

the AY_Ctor constructor does not set the chip frequency:

sc3-plugins/source/AYUGens/AY_UGen.cpp

Lines 49 to 51 in 1c9bb52

	ayemu_set_chip_type(ay, chiptype, NULL);
	//ayemu_set_chip_freq(ay, freq);
	ayemu_set_sound_format(ay, freq, chans, bits);

therefore we fall back to the default frequency of 1773400, defined here:

sc3-plugins/source/AYUGens/AY_libayemu/src/ay8912.c

Line 17 in 1c9bb52

AYEMU_DEFAULT_CHIP_FREQ = 1773400

so, some real-world values of ChipTacts_per_outcount are: (keeping in mind this is all integer division)

1773400/44100/8 = 5
1773400/48000/8 = 4
1773400/96000/8 = 2

now we can start to see where the magic number 110300 in freqtotone comes from; it is very close to 44100*5/2 = 110250. this hints at a general formula for the conversion constant:

sampleRate * chipTacts_per_outcount / 8

and indeed, this formula seems to work :) notes now sound in tune at both 44100 and 48000 Hz sample rates. (though due to the integer calculations, i guess there will always be some error.)

the PR just implements this calculation in AY.freqtotone.

[ahihi]

thanks, good points! ive changed it to take a sample rate instead, and also updated the documentation.