fix(benchmarks): handle EOF when selecting models without -m

[sridipbasu]

Summary

Handle EOFError during interactive benchmark model selection to improve behavior in non-interactive environments.

Previously, running:

kaggle benchmarks tasks run <task>

without -m/--model could fail ambiguously or hang when stdin was unavailable (for example in CI or other non-interactive environments).

This change:

• wraps the interactive input() call in try/except EOFError
• raises a clear ValueError instructing users to pass -m/--model
• preserves existing piped stdin behavior (e.g. echo 1 | ...)

Changes

• Updated _select_models_interactively in kaggle_api_extended.py

• Added tests for:

  • EOF/non-interactive stdin behavior
  • piped stdin compatibility

Test Plan

pytest src/kaggle/test/test_benchmarks_cli.py::TestRun