PLIC refactors and performance improvements

[hawkw]

Currently, the interrupt vector dispatch code uses a linear search over
the interrupt vector array by checking each vector's ID against the
interrupt number being dispatched. This is inefficient; because we
already generate the vector array in order, we can simply index into it,
which is O(1) instead of O(n_interrupts). This commit
changes the existing code to do that.

We can rely on the ordering being correct as it's generated by a const fn that should always output the same order. However, I've also added
debug assertions that the index and ID match, just in case the table
somehow gets corrupted or something. I don't think this is really that
necessary, but it seemed nice to have.

[hawkw]

@jamesmunns I'd love to get your review on this, since I'm hacking up your code. Also, I haven't actually tested this on my D1 yet, so I should make sure it like, works.

[hawkw]

hmm, i flashed this change on my D1 and observed the following:

• the UART doesn't seem to work (no output in crowtty)
• the blue blinky light is solid on rather than blinky

which kind of suggests that interrupt handling is broken.

however, the keyboard and SHARP display do seem to work. going to try flashing with a build from master and making sure we haven't just broken stuff on master too.

[hawkw]

hmm, i flashed this change on my D1 and observed the following:

* the UART doesn't seem to work (no output in crowtty)

* the blue blinky light is solid on rather than blinky

which kind of suggests that interrupt handling is broken.

however, the keyboard and SHARP display do seem to work. going to try flashing with a build from master and making sure we haven't just broken stuff on master too.

Update: nevermind, the "broken UART" was just "i was looking at the wrong USB TTY", and after rebasing onto `main I now see a proper blinkenlight. False alarm!