tp: allow pipelines as subqueries

[LalitMaganti]

Pipelines can only be whole statements or the body of a CREATE PERFETTO TABLE, so nothing written in SQL can use one: not a join, a CTE, a view or a function. That is what stops users like wattson moving off the interval intersection macros.

A pipeline in parentheses now reads like any other subquery, in a FROM clause or as a CTE:

SELECT t.id, p.total
FROM tree t
JOIN (FROM tree |> TREE ACCUMULATE UP SUM(self) AS total) p USING (id)

How it works

• Parser: the grammar hands each such pipeline to syntaqlite's node expander once it is parsed (syntaqlite 0.11, rolled in tp: roll syntaqlite to 0.12.1 #7635). It is compiled there and replaced with SQL reading its serialized plan (tp: run pipelines from their serialized plans #7631), recorded as a rewrite where it was written, just like a macro call. So SQL and pipelines nest either way round, pipelines work in macros and macros in pipelines, and views, functions and triggers need nothing more than they already do.
• Rewrite tree: the parser's macro rewrite builder becomes a RewriteTree covering pipelines as well as macro calls. It is built once per use and never changes. A pipeline's rewrite stands in for the macro calls written inside it (one KeepOutermost rule). A node written inside a macro's expansion is taken as a slice of that expansion, so errors trace back through the call.
• Compiler: a pipeline read by another pipeline is not expanded. The compiler reads it as one more kind of relation, next to dataframes and SQL (CompileRelation), and compiles it into the same plan in a scope of its own. That also makes it usable as an INTERVAL INTERSECTION OF operand.
• Layering: the SQL that reads a plan (__intrinsic_pipeline, the column limit) moves to pipeline_sql.h, out of serialization and the table function.

Refused for now: a pipeline's SQL source reading a function's arguments ($k), with an error when the pipeline is compiled. Nothing binds them there, so they would otherwise silently read as NULL. Supporting them later means passing them to the plan as parameters.

Tests cover:

• pipelines in subqueries, views and functions;
• pipelines in macros and macros in pipelines, including an error tracing back through the macro call;
• pipelines as pipeline sources and as intersection operands;
• the function-argument error;
• SQL → pipeline → SQL → pipeline nesting over several batches: streaming, re-reads from a join, errors from the bottom level, stopping early and interleaved runs.

[github-actions]

🎨 Perfetto UI Builds & Tests

• 174276b (2026-09-27 06:53 UTC): https://storage.googleapis.com/perfetto-ci-artifacts/gh-36300962894-1-ui/ui/index.html
• 9e10352 (2026-09-27 17:18 UTC): https://storage.googleapis.com/perfetto-ci-artifacts/gh-36333281154-1-ui/ui/index.html
• 065123c (2026-09-27 18:21 UTC): https://storage.googleapis.com/perfetto-ci-artifacts/gh-36337697173-1-ui/ui/index.html
• 980bd04 (2026-09-27 19:36 UTC): https://storage.googleapis.com/perfetto-ci-artifacts/gh-36343810122-1-ui/ui/index.html
• b18a926 (2026-09-27 20:17 UTC): https://storage.googleapis.com/perfetto-ci-artifacts/gh-36345305667-1-ui/ui/index.html

[LalitMaganti]

Superseded by #7644 (pipelines as subqueries), with its refactors in #7642 and CTE support in #7643.