fix(postgres): correctly infer nullability for LEFT JOIN rewritten as RIGHT JOIN

[luca992]

PostgreSQL's planner may execute A LEFT JOIN B as a hash join with Join Type: Right to put the smaller relation on the hash-build side.

This is documented behavior:

• [Postgres docs, 14.3 Controlling the Planner with Explicit JOIN Clauses]

  "Most practical cases involving LEFT JOIN or RIGHT JOIN can be
  rearranged to some extent."
  https://www.postgresql.org/docs/current/explicit-joins.html

• [Postgres Pro, "Queries in PostgreSQL: 6. Hashing"]

  "On the physical level, the planner determines which set is the
  inner one and which is the outer one not by their positions in
  the query, but by the relative join cost. ... So, the join type
  switches from left to right in the plan."
  https://postgrespro.com/blog/pgsql/5969673

After the swap the SQL right operand (the nullable side under LEFT JOIN semantics) ends up as the plan's Outer child instead of the Inner child. The old visit_plan only marked Inner children nullable, which on a Join Type: Right plan:

• marked the SQL left operand (always preserved) as nullable, causing spurious Option<T> in macro output for NOT NULL columns
• failed to mark the SQL right operand as nullable, masking real NULLs and panicking at decode time when no LEFT JOIN row matched

Fix threads the parent join type into visit_plan and picks the NULL-fill side based on it:

• Left -> Inner child is nullable (no change)
• Right -> Outer child is nullable (new)
• Full -> both children are nullable (no change)

It also recurses into all child plans, not only when the current node is Left/Right, so nested joins reached through non-join intermediates like Hash are walked.

Does your PR solve an issue?

fixes #3202

Is this a breaking change?

Behavior change yes. The old behavior was wrong inference, not a documented contract.

For queries with a LEFT JOIN that postgres rewrites as a Hash Right Join (driven by pg_statistic cost estimates, fires on production sized data with plan_cache_mode = force_generic_plan which sqlx-macros-core itself sets), generated types change:

• Columns from the SQL left operand that are NOT NULL in their base table go from Option<T> to T. So code handling these unneeded nulls would need to be dropped.

• Columns from the SQL right operand (the LEFT JOIN nullable side) go from T to Option<T>. Code treating these as non-null was already at risk of unexpected null; try decoding as an Option panics at runtime whenever a row had no matching join partner. After this it stops compiling and the field needs to become Option<T>.

The second case is the bigger one. Code passed type checks before only because LEFT JOINs happened to always find a match in test data. After this fix the type system exposes the real nullability.

[die-herdplatte]

I observed an issue with cargo sqlx prepare returning different metadata between a populated and a non-populated DB.
This PR fixes the original discrepancy (which was related to a LEFT JOIN) but opens up another discrepancy elsewhere.

In my case, the query plan on the unpopulated DB contains a root output of the shape ((<expr>)).
During traversal via visit_plan, there's a plan node where parent_nulls_this_side evaluates to true and contains the root output in its own outputs.
The only problem here is that the plan output is represented as (<expr>) (with only one pair of outer parentheses), so the exact match with the root output will fail and the the recorded nullability will remain unknown / None.

The generated query plan for the populated DB is different and doesn't contain such an expression as one of its root outputs.
Here, nullability for the same output will be detected as true.
This results in cargo sqlx prepare --check to fail depending on the state of the reference DB.

[luca992]

@die-herdplatte hmm is there a way you can share for me to reproduce it?