ref(rust-consumers): validate_schema option

[MeredithAnya]

Adds a validate_schema option to the snuba sentry-options namespace so schema validation can be turned off per storage. Leaves the --enforce-schema as is, and makes that take prescedence if it's true

The option is a dict mapping storage name to a bool:

{ "eap_items": false }

• Storages with no entry default to true, so this is a no-op until a storage is explicitly listed.
• false skips validation and skips loading the schema entirely.
• --enforce-schema wins: those consumers always validate and DLQ, regardless of the option.

The option is read once when the consumer builds its processing strategies, then carried as a plain bool — never looked up per message, which is the cost this is meant to remove. So changing it requires a consumer restart or a rebalance. Both the Rust processor paths and the hybrid Python-transform path are covered.

Tests cover the default-on behavior, per-storage opt-out, and that --enforce-schema still validates when the option says false.

[tryangul]

LGTM. Q: Do we need to update the python side as well for consistency? Also, what is the risk here? I didn't notice any problems running without validation in s4s2, personally, but maybe there's more 🌶️ in prod.

[MeredithAnya]

LGTM. Q: Do we need to update the python side as well for consistency? Also, what is the risk here? I didn't notice any problems running without validation in s4s2, personally, but maybe there's more 🌶️ in prod.

This is mainly to test improves on the eap consumer, so I wasn't really thinking about the python side here, (updated the title to be more accurate)

Haven't looked too far but seems like we are doing validation for the few consumers we have

snuba/snuba/consumers/consumer.py

Line 478 in 3170625

validate_sample_rate = (