Reindexer: Mock reindex function to reduce parallel contention

[brandur]

This one's aimed at fixing an intermittently failing test case:

https://github.com/riverqueue/river/actions/runs/24322049420/job/71009998068?pr=1208

--- FAIL: TestReindexer (0.00s)
    --- FAIL: TestReindexer/ReindexesConfiguredIndexes (10.07s)
        reindexer_test.go:219: Reusing idle postgres schema "maintenance_2026_04_13t01_54_14_schema_04" [user facing: "maintenance_2026_04_13t01_54_14_schema_04"] after cleaning in 26.436456ms [4 generated] [7 reused]
        test_signal.go:95: timed out waiting on test signal after 10s
        logger.go:256: time=2026-04-13T01:54:27.581Z level=INFO msg="maintenance.Reindexer: Signaled to stop during index build; attempting to clean up concurrent artifacts"
        riverdbtest.go:293: Checked in postgres schema "maintenance_2026_04_13t01_54_14_schema_04"; 1 idle schema(s) [4 generated] [10 reused]
    --- FAIL: TestReindexer/ReindexesMinimalSubsetofIndexes (10.14s)
        reindexer_test.go:183: Reusing idle postgres schema "maintenance_2026_04_13t01_54_14_schema_01" [user facing: "maintenance_2026_04_13t01_54_14_schema_01"] after cleaning in 28.042877ms [4 generated] [10 reused]
        test_signal.go:95: timed out waiting on test signal after 10s
        reindexer_test.go:211:
                Error Trace:	/home/runner/work/river/river/internal/maintenance/reindexer_test.go:211
                Error:      	Should be false
                Test:       	TestReindexer/ReindexesMinimalSubsetofIndexes
        logger.go:256: time=2026-04-13T01:54:28.444Z level=INFO msg="maintenance.Reindexer: Signaled to stop during index build; attempting to clean up concurrent artifacts"
        riverdbtest.go:293: Checked in postgres schema "maintenance_2026_04_13t01_54_14_schema_01"; 1 idle schema(s) [5 generated] [24 reused]
FAIL
FAIL	github.com/riverqueue/river/internal/maintenance	18.764s

I'm diagnosing with Claude's help here, but what appears to be happening
is that although a reindex operation in Postgres is often fast, it is
still a heavy operation, and can slow down even further when there's a
lot of concurrent activity hammering a database.

Many reindexer test cases run in parallel, and it appears that was
happening here is that we got a reindex that exceeded our maximum
timeout of 10x. We have some evidence this during the test run from the
runtime of 10.07s and the line:

Signaled to stop during index build; attempting to clean up concurrent artifacts

Here, I'm putting forward a solution proposed by Claude, which is to
mock out the reindex operation, especially where we have a number of
reindexes going in parallel. The tests ReindexOneSuccess and
ReindexSkippedWithReindexArtifact still put load on the real reindex
operation, so we're not going full mock here, and should see our
intermittency considerably reduced while still being confident that
everything still works.