feat: synchronous remote index dropping

[usamoi]

• closes bug: error report #388 (task 2)
• closes feat: synchronous remote index dropping #340 (task 1, 2, 3)
• closes feat: refactor async task executor #260 (task 1, 3; a global thread pool may not be implemented)
• closes improvement: Separation of backend services  #307 (algorithms and index are split off the service crate)

---

• Most code in service is moved to crates/index and crates/{algorithm name}. We can improve and test these algorithms independently without DB stuffs.
• We run crates/{algorithm} in a thread pool while indexing. It's a copy of rayon and we support rayon::check to check cancellation point now.
• Handle is refactored to take advantage of system identifier and the index oid. Based on it, we reimplemented drop logic: we drop remote index in a object_access_hook.
• Old prefilter code is removed.
• This PR introduces jemalloc. The new pgrx release makes it possible.
• Added a toml format check to ci.

[VoVAllen]

Please update the details to be more specific about the change since this PR is huge

[VoVAllen]

ZomboDB generates a UUID to use as the backing Elasticsearch index name, but also assigns an alias in the form of
database_name.schema_name.table_name.index_name-index_oid. "index_oid" is the Postgres catalog id for the index from
the "pg_class" system catalog table.

Shall we use similar approach?

Ref: https://github.com/zombodb/zombodb/blob/1416c99a4885f1cfe5e7bd86b935e89e6d6ba431/INDEX-MANAGEMENT.md?plain=1#L20C1-L23C1

[usamoi]

ZomboDB generates a UUID to use as the backing Elasticsearch index name, but also assigns an alias in the form of
database_name.schema_name.table_name.index_name-index_oid. "index_oid" is the Postgres catalog id for the index from
the "pg_class" system catalog table.

Shall we use similar approach?

Ref: https://github.com/zombodb/zombodb/blob/1416c99a4885f1cfe5e7bd86b935e89e6d6ba431/INDEX-MANAGEMENT.md?plain=1#L20C1-L23C1

I think it's not necessary.