sql: add EXPLAIN (VEC) output to the stmt bundle

[yuzefovich]

sql: delay saving of flows till after the setup

This commit reorders the code that deals with the flow specifications
a bit. Previously, we would call saveFlows right after generating flow
specifications and we would release the resources back to their pools
once the setup request for each flow has been sent out. However, the
follow-up commit will include the generation of EXPLAIN (VEC, VERBOSE)
diagram into the statement bundle, and to get there we need to delay
both of those actions until we know that the flows are vectorized. As
a result, now saveFlows is called after setupFlows (when we know
whether they were vectorized or not) and the release of the flow specs
is deferred.

Release note: None

sql: add EXPLAIN (VEC) output to the stmt bundle

This commit refactors the code of EXPLAIN (VEC) to be reused to add
(VEC) and (VEC, VERBOSE) flavours to the statement bundle. The code
to print the diagram is moved into colflow package, but other changes
are minor.

The addition of these diagrams is performed on the best effort basis
(meaning that if an error is encountered during populating the diagram,
the error is ignored) because in some edge cases the errors are expected
to occur.

Fixes: #47940.

Release note (sql change): Statement diagnostics bundles now contain
output of EXPLAIN (VEC) and EXPLAIN (VEC, VERBOSE) commands for the
statements.

[cockroach-teamcity]

This change is 

[yuzefovich]

Confirmed manually that it works:

[RaduBerinde]

Very cool!

The lists of files in TestExplainAnalyzeDebug need updating.

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @yuzefovich)

---

pkg/sql/distsql_physical_planner.go, line 700 at r2 (raw file):

		var explainVecVerbose []string
		if planner.instrumentation.collectBundle && planner.curPlan.flags.IsSet(planFlagVectorized) {
			flowCtx := newFlowCtxForExplainPurposes(p, planner, &planner.extendedEvalCtx.DistSQLPlanner.rpcCtx.ClusterID)

Given that a panic in this code could be very bad, consider adding a panic catcher in this function (you can copy the block in emitExplain()).

---

pkg/sql/explain_bundle.go, line 324 at r2 (raw file):

			}
			if len(d.explainVec) > 0 {
				b.z.AddFile(fmt.Sprintf(filenameFormat, ""), strings.Join(d.explainVec, "\n"))

[nit] The filenameFormat logic is hard to follow (especially since we're intermixing it with other Sprintfs). I'd do something simpler:

extra := ""
if len() > 1 {
  extra = fmt.Sprintf("-%d-%s")
}
AddFile(fmt.Sprintf("vec%s.txt", extra))
AddFile(fmt.Sprintf("vec%s-v.txt", extra))

[blathers-crl]

Thank you for updating your pull request.

Before a member of our team reviews your PR, I have some potential action items for you:

• We notice you have more than one commit in your PR. We try break logical changes into separate commits, but commits such as "fix typo" or "address review commits" should be squashed into one commit and pushed with --force
• When CI has completed, please ensure no errors have appeared.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is otan.}

[yuzefovich]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde)

---

pkg/sql/distsql_physical_planner.go, line 700 at r2 (raw file):

Previously, RaduBerinde wrote…

Given that a panic in this code could be very bad, consider adding a panic catcher in this function (you can copy the block in emitExplain()).

We already had the vectorized panic catcher in ExplainVec, and I extended it to the whole function body instead of introducing a panic catcher here.

[yuzefovich]

TFTR!

bors r+

[craig]

Build succeeded:

• GitHub CI (Cockroach)

[yuzefovich]

Thoughts on backporting this to 21.1? I think it should be pretty safe and somewhat beneficial.

[jordanlewis]

I think this is a good idea to backport.