[PATCH] Add pg_current_vxact_id() function to expose virtual transaction IDs

Incremental Update: New Review from Solai

A new participant (Solai) has posted a review/testing report that adds modest technical substance to the discussion, primarily reinforcing existing arguments with explicit testing observations rather than introducing new technical ground.

What's New

Solai's message provides two minor technical observations that weren't explicitly articulated before:

1. pg_locks deduplication issue: The observation that a single transaction can appear multiple times in pg_locks.virtualtransaction (once per lock held) means the naive SELECT virtualtransaction FROM pg_locks WHERE pid = pg_backend_pid() LIMIT 1 workaround isn't just expensive — it's semantically fragile. Without DISTINCT or LIMIT, it returns duplicates. This was implied but not stated explicitly in prior discussion.

2. VXID stability across transaction states: Explicit confirmation through testing that VXID is present and stable from the start of any transaction (including read-only), whereas pg_current_xact_id_if_assigned() returns NULL until a write occurs. This was understood conceptually but Solai frames it as "the only identifier that exists uniformly for all transactions" — a slightly stronger framing of the completeness argument.

Assessment

This is essentially a "+1 with testing confirmation" message. It does not introduce new technical arguments, propose design changes, or identify issues. It adds weight to the community consensus side of the equation (relevant given Michael Paquier's "outvoted" framing) but does not advance the patch technically. No new patch version was posted.