chain/ethereum: Fix dropped block trigger when once and polling filters match same block (#6530)

* chain/ethereum: Fix dropped trigger when once and polling filters match same block

blocks_matching_polling_intervals used find_map over polling_intervals,
which short-circuits after the first matching entry. Since a single
(start_block, interval) tuple satisfies at most one of the once/polling
conditions, a block that should emit both Start and End only got one.
HashSet iteration order made which trigger survived non-deterministic
across process restarts, causing PoI divergence.

Replace with two independent any() scans so both conditions are
evaluated across all entries, matching parse_block_triggers on the
BlockFinality::NonFinal (reorg-window) path.

* chain/ethereum: Extract block_trigger_types_from_intervals helper

Both blocks_matching_polling_intervals (JSON-RPC path) and
parse_block_triggers (Firehose path) now compute the same
once/polling match logic. Pull it into a shared pure function
so the two paths cannot drift out of sync.

* chain/ethereum: Add tests for block_trigger_types_from_intervals

Cover the determinism-critical cases the new helper guards against:
once-only, polling-only, once + polling sharing a start_block,
cross-datasource collision where a polling schedule lands on another
datasource's once block, and the empty/no-match cases.