feat(memify): add status endpoints for async runs

[slime-turing]

Summary

This PR adds a small status API for async memify runs.

You can now:

• call GET /api/v1/memify/status to fetch the latest memify run for a dataset
• call GET /api/v1/memify/status/{pipeline_run_id} to track one specific background run

It also includes focused API tests, a short example, and docs that explain why pipeline_run_id is the right polling key for async memify.

A couple of details worth calling out

While working through this, I tightened two things that seemed important before opening the PR:

• the status endpoint no longer returns raw persisted input payload from run_info.data
• name-based dataset lookups now fail fast if more than one readable dataset shares the same name, and ask the caller to retry with dataset_id

Why this shape

The endpoint is meant to be cheap to poll and predictable to use:

• it reuses the existing pipeline run lookup path
• it keeps the exact-run flow centered on pipeline_run_id
• it avoids returning unnecessary payload data in the status response

Validation

I checked this with focused unit tests and with a live HTTP probe against the branch build.