RVFI CSR support

Nobody has claimed this yet.

Is there a reason Ibex's RVFI interface doesn't support CSR writes? I'd like to add it so we can trace CSR writes.

Note there appear to be two different approaches:

1. [Give each supported CSR a wire.](https://github.com/SymbioticEDA/riscv-formal/blob/master/docs/rvfi.md#control-and-status-registers-csrs)
2. [Use a vector of wires](https://docs.openhwgroup.org/projects/cv32e40s-user-manual/en/latest/rvfi.html#:~:text=instructions%20have%20retired.-,CSR%20Signals,-To%20reduce%20the). Hard to tell what each index corresponds to.
3. [Use a 4096-element vector of wires!](https://github.com/riscv-verification/RVVI/blob/0043bdf0313670f3e1d6bac3af2d88b52de459d6/source/host/rvvi/rvviTrace.sv#L91)

That last one seems pretty mad. I think the first options is probably best given the relatively small number of CSRs that Ibex has. I'd probably put them in an `interface` to save wiring insanity. Something like:

```
struct rvfi_csr {
logic [31:0] wdata;
logic wvalid;
}

interface rvfi_csrs {
rvfi_csr mstatus;
rvfi_csr mcause;
...
}
```

I don't think there's any need to record CSR reads. I'm not aware of any with read side effects (`seed` almost is but you aren't allowed to only read it). Does that sound reasonable?

Start with the linked RVFI CSR documentation and compare its named-wire and vector approaches with Ibex's existing RVFI interface. The work is done when the CSR write representation is agreed and CSR writes can be traced through the RVFI interface; the issue names no Ibex files or tests to begin with.

Written by the indexing model from the issue text.