Fixing an issue where valid HID responses are discarded

[infinityis]

I have observed that occasionally VIA will not complete the Authorize Device process. Also, for a larger keyboards especially, the Test Matrix feature will consistently lock up and stop responding after a few seconds (whereas disabling "Test Matrix" continues to work just fine). After a bit of in-browser debugging, I learned that the call to fastForwardGlobalBuffer discards HID responses received prior to whenever read was called.

It turns out that due to the use of async/await, there can be some delay between the write function call and the read function call. What was happening was that the device had already responded quickly between the time when write was called and the time when read was called. Whenever the device responds before read is called, that response is then erroneously discarded by the fast forward function. VIA is then stuck waiting for a response that will never arrive (because it already did arrive, and was discarded), causing Test Matrix to stop working, and/or causing the authorization process to halt.

This PR makes it so that fastForwardGlobalBuffer only discards messages prior to the previous call to write instead of discarding messages prior to when read is called.

Note that these changes were tested by editing the code through browser developer tools by enabling overrides and thus running it in-browser. The changes in this PR match what I did in-browser and it looks correct to me, but I haven't actually run/tested the committed changes as-is simply because I haven't set up an environment to run/host VIA independently. That being said, I can confirm these when I made these same changes in-browser, they worked and actually fixed the issues I was experiencing.

Edit: It should also be noted that the problems I mentioned would only show up when the CPU load was higher (i.e. lots of tabs open in the browser, or on older computers).