Fix flaky compactor tests: ignore user index update loop errors

[CharlieTLe]

What this PR does:

Fixes a flake in pkg/compactor where background log output non-deterministically breaks the tests
that assert the compactor's complete log set.

userIndexUpdateLoop runs in the background of every sharding-enabled compactor test and ticks every
UsersScanner.UpdateInterval, which prepare() sets to 100 * time.Millisecond. Each tick calls
ownUser(userID, true), which consults the ring, so any tick that lands before the lifecycler is
ACTIVE — or after it starts LEAVING at shutdown — logs:

level=error component=compactor msg="failed to check if compactor owns updating user index" err="at least 1 live replicas required, could only find 0 - unhealthy instances: 1.2.3.4:0"

The 7 assertions in compactor_test.go and 7 in compactor_paritioning_test.go that compare the full
log output with ElementsMatch then fail with elements differ, depending purely on how many ticks
fell inside the test window.

removeIgnoredLogs already suppresses the other two lines this loop can emit
(successfully updated user index, context timeout, exit user index update loop), but not these
two. Both carry a variable err= payload, so they need a regex rather than an exact map entry —
likely why they were missed when the others were added.

This is a test-synchronisation bug, not a product bug. The errors are legitimate signal in
production, so the fix is in the test helper rather than the loop.

Which issue(s) this PR fixes:
Fixes #7826

How this was verified:

The flake is timing dependent and does not reproduce locally, so this PR adds
TestRemoveIgnoredLogs_UserIndexUpdateLoop to cover the helper directly. I confirmed it fails
without the fix:

--- FAIL: TestRemoveIgnoredLogs_UserIndexUpdateLoop (0.00s)
    Error: Should be empty, but was [level=error component=compactor msg="failed to check if
    compactor owns updating user index" err="at least 1 live replicas required, could only find 0 -
    unhealthy instances: 1.2.3.4:0" level=error component=compactor msg="failed to update user index"
    err="mocked error"]

and passes with it. The full package also passes under the CI configuration:

$ go test -race -tags "netgo slicelabels" -count=1 ./pkg/compactor/...
ok  	github.com/cortexproject/cortex/pkg/compactor	115.057s

Notes:

• Also ignores failed to update user index from the same loop. Not yet observed in CI, but it has
  identical exposure.
• Not covered by Flaky pkg/compactor tests: sub-second Poll budgets and polling a different signal than the assertion #7794 / Fix flaky pkg/compactor tests #7796, which fixed a different mechanism in this package (short Poll
  budgets, failing as expected 1, got 0). This one is an assertion-content race failing as
  elements differ.
• No CHANGELOG.md entry: test-only change with no user-facing impact, consistent with Fix flaky pkg/compactor tests #7796.

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]
• docs/configuration/v1-guarantees.md updated if this PR introduces experimental flags