Fix issues with input detection due to overloaded ring buffer and fix buffer size being incorrect

[kalaposfos13]

Adds an axis noise filter and only updates outputs if a change happened to that output, meaning that there are now both less axis movement event spam, and now every input event only updates one output, instead of updating all of them with all but the updated one holding duplicate data from the previous entry, which was making the ring buffer holding input events being overloaded in two different ways. These features were once in my main remap PR, but I had to revert them, although it looks like whatever was the issue that led to that no longer applies, as from my testing, I couldn't find any significant regressions, but just to be safe, I'll open this as a draft and wait until it's tested on more games.

[sam20112011]

This PR fix this issue for me: #2340

[Hermiten]

Does this fix input lag visible on these games https://github.com/shadps4-emu/shadps4-game-compatibility/issues?q=is%3Aissue%20state%3Aopen%20label%3A%22input%20lag%22 ?

[ElBread3]

Tested Minecraft, this PR appears to have gotten rid of menu input lag, but not any game input lag.

[kalaposfos13]

I think I found a possible fix for Minecraft's (and maybe for other games') input lag, which is reducing the buffer size to 32, which is the number that's seen in Ghidra. It can be tested here: https://github.com/kalaposfos13/shadPS4/tree/mc-input-lag-hack

[kalaposfos13]

Since this is a draft, and the ring buffer size change is at least partially related to this PR anyway, I figured I might as well add the Minecraft fix to here.

[Hermiten]

I think you can and if there is a big problem, you will split it
Now we need some tests on games with and without this problem to be sure everything is good