Add first-class skip_pg_for_brainstore_objects variable for no-PG mode

[Alex Soldatchenko (soldatchenko)]

Adds skip_pg_for_brainstore_objects as a first-class Terraform variable, replacing the need for customers to manually configure three separate brainstore_extra_env_vars entries.

When set, a single variable fans out to:

• SKIP_PG_FOR_BRAINSTORE_OBJECTS on the API Handler Lambda
• BRAINSTORE_WAL_USE_EFFICIENT_FORMAT on the API Handler Lambda (auto-enabled)
• BRAINSTORE_ASYNC_SCORING_OBJECTS on all Brainstore EC2 nodes
• BRAINSTORE_LOG_AUTOMATIONS_OBJECTS on all Brainstore EC2 nodes

This mirrors the Helm chart's skipPgForBrainstoreObjects single-knob pattern.

Accepted values: "" (disabled), "all", "include:<type>:<uuid>,...", "exclude:<type>:<uuid>,...".

Open question: WAL_USE_EFFICIENT_FORMAT auto-coupling

This PR auto-enables BRAINSTORE_WAL_USE_EFFICIENT_FORMAT = "true" wheneverskip_pg_for_brainstore_objects is set (matching Helm chart behavior). However, it'sunclear whether this setting has the same sequencing constraint as BRAINSTORE_WAL_FOOTER_VERSION,where old Brainstore nodes can't read the new format during a rolling upgrade.

If it does, auto-enabling it with no-PG could be unsafe for upgraded deployments (fresh deploys are unaffected). Waiting on confirmation before removing the draft status.

Options under consideration:

• Keep as-is (matches Helm, assumes the coupling was validated)
• Gate it behind brainstore_wal_footer_version being set (safer for upgrades)
• Decouple it into a separate explicit variable