Rename ExecuteOptions.unique_id to loadshed_valve_id

[bgwines]

Background / Why?

The Snake load shedder keys its per-valve self-contention layer on a field of ExecuteOptions, but that field was named unique_id while both the Snake RFC and the loadshed package call the concept a "valve ID". The mismatch made the wiring confusing to follow — call sites read GetUniqueId() and immediately passed the result into Acquire(ctx, valveID, ...), so a reader had to know that unique_id was the valve ID. The RFC specifies the field should be named loadshed_valve_id.

This renames the proto field unique_id → loadshed_valve_id. Field number 22 is unchanged, so the wire format stays backwards-compatible; only the Go symbol and JSON name change. ExecuteOptions is forwarded from vtgate to vttablet wholesale, so there is no per-field passthrough to update — the only readers are the two Snake acquire sites (QueryExecutor.getConn and TxPool.Begin), which now call GetLoadshedValveId().

While here, the "an empty valve ID is valid" explanation that previously sat at both call sites is moved into Snake.Acquire's doc comment, where the contract actually lives, so it is not duplicated and is visible to any future caller.

The generated .pb.go files were regenerated with the same protoc-gen-go version already used for the checked-in code (v1.36.11), so the diff is limited to the rename rather than a toolchain-version reformat.

Testing

Updated the existing Snake unit tests to use the renamed field (LoadshedValveId in ExecuteOptions literals) and ran go test for go/vt/vttablet/tabletserver and go/vt/vttablet/tabletserver/loadshed. go build ./go/vt/... confirms the rename is consistent across the tree (the only non-generated readers are the two acquire sites).

AI disclosure: Claude Code assisted with development. Every line of code was either written by or carefully reviewed by me :)