RISC-V Extensions update including 29 extensions to stabilize

[a4lg]

This commit directly corresponds to rust-lang/rust#145948.
See that proposal for criteria of 29 new extensions.

Notes (for the Rust reference)

• Updated existing a and c to imply its subset(s).
  • a equals zalrsc and zaamo combined (imply both from a).
  • The C extension always implies the Zca extension and conditionally implies Zcf and/or Zcd extensions (on RISC-V) but
    c (Rust target feature) only implies zca as both zcf and zcd are not to be stabilized on this proposal.
• For each extension, linked to the first document containing it with the ratified state.
  For instance, all extensions present in the RVA23 profiles link to (3) (despite that they are present in (2)) because they were not ratified at the time when (2) is published.

References

1. RISC-V Instruction Set Manual (version 20250508)
   https://github.com/riscv/riscv-isa-manual/tree/20250508
2. RISC-V Profiles (version 1.0 - RVA23 profiles were not ratified at the time)
   https://github.com/riscv/riscv-profiles/tree/v1.0
3. RISC-V Profiles (RVA23/RVB23-ratified version)
   https://github.com/riscv/riscv-profiles/tree/rva23-rvb23-ratified

[a4lg]

Possible Consideration

Some platform behavior extensions are hard to make a short, concise summary.

Do we prefer to keep long(er than usual) descriptions? Or, do we just link to the RISC-V Profiles documentation and let it explain?

My Current Position

I prefer the latter if allowed but mdbook design is confirmed not broken so the current approach (long descriptions) is okay for me.

[a4lg]

Ping. Can someone review this PR?

[a4lg]

Non-stabilization part is separated as #2002.

Submission v2 (zacas is added) of this PR is a superset of #2002.

[a4lg]

cc: @Amanieu, @RalfJung

[rustbot]

This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[a4lg]

Rebased and made some clarification in the commit message (the same as the edited initial comment).