NES: FDS Accuracy Improvements

[TakuikaNinja]

This PR aims to improve the FDS accuracy based on ongoing research: https://www.nesdev.org/wiki/User:TakuikaNinja/FDS_Registers

I'm not as familiar with the audio side of things, so any help there would be appreciated.

Working

• $4023.D0 only affects writable disk registers, forces certain values for $4025 and $4026
• $4023.D1 only directly affects $4080 (volume envelope), $4085 (mod counter), and $4088 (mod table write), apply instant volume mute for now
• External connector: Mask $4026 writes to 7 bits, and force bit 7 = 1 for $4033 reads (i.e. always return good battery/power status)
• $4030 reads do not clear byte transfer flag/disk IRQs
• Move byte transfer flag to $4030.D7 (eventually needed by Tonkachi Editor)
• A more faithful modulation implementation, based on Implement nonlinear FDS DAC and updated FDS audio emulation SourMesen/Mesen2#33 (minus non-linear DAC)
• $4090-$4097 reads (debug audio registers), which required the above implementation

Needs testing/fixing

• Modulation issues in FDS-Audio-Registers
• Modulation issues in Bio Miracle Bokutte Upa (e.g. the invincibility theme when grabbing a bell item, underground theme in 1-2)
• Test for regressions in commercial games + NSFs, especially the modulation

Not addressing

• Add remaining disk registers to register view
• Register peeking
• Actual audio reset state when $4023.D1 = 0 (research ongoing) + rename _soundRegEnabled to reflect its purpose
• $4095 upper nybble
• DRAM refresh watchdog ($4030.D1)
• End of head behaviour ($4030.D6)
• Byte transfer flag behaviour ($4030.D7)
• Non-linear DAC + volume PWM + filtering (probably not worth the performance hit for a minuscule audio improvement)

Test programs

• FDS-Mirroring-Test (Passes all tests on latest version)
• FDS-4023-Test
• FDS-Audio-Registers

[TakuikaNinja]

Turns out the PR I was referencing for the modulation had missed the fact that the mod accumulator is halted while $4087.D7 = 1. FDS-Audio-Registers sounds correct now.

Bio Miracle Bokutte Upa still sounds weird compared to recordings I found on TCRF, however, so I'll have to double check against my Twin Famicom. Maybe the mixing resistor swap will make it easier to discern! 😛

[TakuikaNinja]

Scaling back the scope of this PR a bit - the remaining register view changes will go into a separate PR, and register peeking will have to wait until more accuracy improvements have been made.

Note: The audio debug registers ($4090-$4097) have already been added to the register viewer under an appropriate heading.