A prototype of much faster langfuse ingestion

[vaibhavbetteromics]

What does this PR do?

Langfuse ingestion is quite slow, because the ingestion code path makes read/write SQL queries for each observation that's received.

This PR attempts a different version of the ingestion code path that reads and writes to the DB in batches.

The performance improvement is significant.

Before:
POST request with 260 events completes in 5s

Running on 100 gleason docs:

Processed 100 documents in 0m4s
POST http://localhost:6007/api/public/ingestion took 3.7955631249933504s with 1007.714KB 307 events
POST http://localhost:6007/api/public/ingestion took 8.681478124985006s with 1446.973KB 314 events
POST http://localhost:6007/api/public/ingestion took 9.812697500048671s with 825.628KB 335 events
Langfuse flush took 0m19s

After:

POST request with 260 events completes in 873ms

Running on 100 gleason docs:

POST http://localhost:6007/api/public/ingestion took 1.0489602909656242s with 920.933KB 264 events
POST http://localhost:6007/api/public/ingestion took 0.7444410839816555s with 1653.679KB 400 events
POST http://localhost:6007/api/public/ingestion took 0.5013136669876985s with 705.703KB 292 events
Processed 100 documents in 0m4s
Langfuse flush took 0m0s

Type of change

• Bug fix (non-breaking change which fixes an issue)
• Chore (refactoring code, technical debt, workflow improvements)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactor (does not change functionality, e.g. code style improvements, linting)
• This change requires a documentation update