fix(storage): make the cassandradb provider work on apache cassandra

[lakhansamani]

Answering "did we run a local Cassandra test?" — we never had. cassandradb was missing from the storage suite's backend list, so the provider that serves both cassandradb and scylladb was only ever exercised through Scylla. Wiring it up and pointing it at cassandra:4.1 fails on the first run.

What was broken

Three queries use CQL LIKE. Apache Cassandra accepts LIKE only against a SASI index and rejects it against the plain secondary index this provider creates:

InvalidRequest: LIKE restriction is only supported on properly indexed columns.
event_name LIKE 'test_event%' is not valid.

ScyllaDB accepts the same statements, which is exactly why this survived.

Method	Effect on Apache Cassandra
GetWebhookByEventName	errors — webhook dispatch by event name never fires
DeleteAllSessionTokensByUserID	errors — logout/session revocation fails
DeleteSessionTokensByNamespace	errors — same

All three already carried ALLOW FILTERING, i.e. a full table scan, so the index was never doing any work. Moving the comparison into Go changes nothing but where it runs — no new scans, no behaviour change on Scylla.

A second bug, on both engines

DeleteAllSessionTokensByUserID matched "%" + userId + "%" — unanchored. A delete for 42 also took out tokens stored for 142 or provider:427. The SQL provider was anchored long ago (user_id = ? OR user_id LIKE '%:' || ?) with a comment explaining precisely this; the Cassandra provider kept the defect, because the assertion covering it sat inside testSQLCRUDCorrectnessFixes, which only runs for SQL backends.

It is now Anchored Session Token Delete, running on every backend. Mutation-verified: restoring the unanchored match fails it on ScyllaDB with substring-matching other user's session token must survive.

Also

Webhook_Operations indexed fetchedByEventName[0] after only an assert.NotNil. An empty-but-non-nil slice panics there, and a panic in one backend's subtest aborts the run for every backend queued behind it — which is how the first Cassandra run ended, hiding whatever came after. Now require.NotEmpty.

Coverage

• cassandradb added to allDBTypes and getTestDBConfig, on port 9142 — deliberately not 9042, so TEST_DBS=scylladb,cassandradb can't bind both types to one engine and silently test Scylla twice.
• make test-cassandradb, same capture-status-then-teardown shape as the other single-DB targets.
• cassandradb added to test-all-db, test-docker-up, test-cleanup, wait-for-test-dbs.sh, and the AGENTS.md table (7 DBs → 8).

Verification

• make test-cassandradb equivalent against cassandra:4.1: all 26 subtests pass (was: panic at subtest 9).
• TEST_DBS=scylladb full storage suite: pass.
• TEST_DBS=sqlite go test -p 1 ./...: pass. go build ./..., go vet ./..., gofmt clean.

---

Second commit: the storage suite assumed an almost-empty table

Chasing #813's remaining failure, I ran the scylladb storage suite in a loop and hit 15 consecutive full-suite failures — on main and on pre-#818. Neither was the LWT bug; both were the test's own fault, and both are the "flaky ScyllaDB storage tests" shape #813 reports.

1. ListUsers first-page assumption. User_Operations asserted the just-created user appears in ListUsers(Limit: 10, Offset: 0) unfiltered. That only holds while the table is nearly empty. At 463 accumulated rows it failed every run with created user should appear in ListUsers — indistinguishable from a write that never landed. Now filtered by the user's own email, which keeps both assertions (ListUsers works; the created user carries its password) and makes them independent of how much data is already there.

2. Cascading pollution from fixed phone literals. The failure above aborts before the cleanup at the bottom of the subtest, leaving +1234567891 in the table. Every later run then failed earlier, with user with given phone number already exists — one failure turning into a different, misleading one. Both numbers are unique per run now, like the emails already were.

Invisible in CI, which gets a fresh container per run. Deterministic locally.

Verification

Same ScyllaDB container, keyspace holding 463+ users throughout:

	before	after
full storage suite × 15	15/15 failed	0/15 failed
User_SCIM_Fields / Org_Domain_Operations / Session_Token_Operations × 100	0 failed	0 failed

Plus the Apache Cassandra suite (26/26) and TEST_DBS=sqlite go test -p 1 ./..., both green.

[lakhansamani]

Superseded by #824, which carries this change unmodified alongside the other three commits in one PR — no merge-order dependency between them.