fix: Drop the py3-min BASE_IMAGE override from the inferentia SDK build

[mc-nv]

What does the PR do?

• The inferentia QA script overrode Dockerfile.sdk's BASE_IMAGE with
  tritonserver:${UPSTREAM_CONTAINER_VERSION}-py3-min, pinning the image being
  retired and contradicting that Dockerfile's own default.
• Removed rather than repointed: the replacement couples train to CUDA version,
  which UPSTREAM_CONTAINER_VERSION cannot express.

Where should the reviewer start?

• inferentia/qa/setup_test_enviroment_and_test.sh — the removed --build-arg
  line in the SDK docker build.

Related PRs:

• fix: Build on cuda-dl-base devel and ship on inference-runtime server#8994
• fix: Swap the trt_llm_backend base image default off py3-min tensorrtllm_backend#863

Related Issues:

• Resolves: TRI-1930

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Removes a Docker build argument from the SDK test setup.

The PR appears safe to merge based on the reviewed change.

Summary

The PR removes the Inferentia QA script’s explicit SDK BASE_IMAGE override, leaving image selection to the checked-out server repository’s Dockerfile.sdk.

• The locally built server remains the QA container’s runtime base.
• The SDK image supplies client artifacts to that QA container.

_{Reviews (1) · Last reviewed commit: "fix: Drop the py3-min BASE_IMAGE overrid..."}