feat(sagemaker): support batch-transform

[deepankarm]

Goals

PR in continuation to #6046.

Sagemaker still calls the POST /invocations route during batch-transform but with different content-type & body. To enable both application/json (for inference) and text/csv (for batch-transform, I removed the pydantic model from the path signature, rather use fastapi's Request class and evaluate the body according to the content-type.

For batch-transform,

• Expect text/csv or application/csv content type.
• Validate if the format is correct.
• Convert each row in the csv to input_model (user-defined model in the Executor route).
• Invoke the route for all inputs together and return the response. This makes sure the values appended to the output csv by sagemaker has a similar format as inference.

Gotchas

• Sagemaker doesn't support CSV with headers. We follow the same restriction.

• Rows in the input csv should have all fields defined in the input type (including optional fields like id)

  from docarray import BaseDoc
  
  class TextDoc(BaseDoc):
      text: str

  Following inputs are valid

  1,abcd
  2,efgh
  3,ijkl
  4,mnop
  5,qrst
  6,uvwx
  7,yzab
  8,cdef
  9,ghij
  10,klmn
  

  Following CSV is invalid.

  abcd
  efgh
  ijkl
  mnop
  qrst
  uvwx
  yzab
  cdef
  ghij
  klmn
  

End-to-end tests

I've manually tested the Executor after pushing a model to S3, the Executor image to ECR & running batch-transform jobs for both SingleLine and MultiLine records.

---

• check and update documentation. See guide and ask the team.

[codecov]

Codecov Report

Attention: 46 lines in your changes are missing coverage. Please review.

Comparison is base (4ea8bb5) 76.93% compared to head (9ff2d57) 76.85%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6055      +/-   ##
==========================================
- Coverage   76.93%   76.85%   -0.09%     
==========================================
  Files         145      145              
  Lines       13889    13916      +27     
==========================================
+ Hits        10686    10695       +9     
- Misses       3203     3221      +18     

Flag	Coverage Δ
jina	76.85% <11.53%> (-0.09%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
jina/orchestrate/deployments/__init__.py	81.67% <37.50%> (-0.05%)	⬇️
jina/serve/runtimes/worker/http_sagemaker_app.py	0.00% <0.00%> (ø)

... and 7 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[JoanFM]

can u add tests with docker?

[deepankarm]

can u add tests with docker?

Added - a7faa6f