chore(connector): remove stale load and benchmark helpers

[GentleCold]

Summary

What does this PR do? (1-3 bullet points)

• Remove stale connector batch-load helpers that were superseded by the request-level load dispatcher.
• Remove dead benchmark correctness, staging, scheduler, and server helpers found during the ponytail repo complexity audit.
• Drop tests that only exercised removed compatibility paths while keeping dispatcher/request-level coverage intact.

Motivation

This cleanup reduces the connector surface area after the prefix load staging pipeline moved to the request-level dispatcher. Keeping both the old batch pipeline and the active dispatcher path made the load code harder to reason about and preserved test-only production helpers.

Test plan

How was this tested?

• Unit tests added / updated
• PYTHONHASHSEED=0 pytest -q -m "not integration" --ignore=tests/integration --ignore=tests/connector/test_daser_connector.py --ignore=tests/connector/test_gds_transfer.py tests/ passes
• Integration tests run, or marked N/A with reason
• Benchmark / e2e smoke run, or marked N/A with reason
• pre-commit run --all-files passes

Validation performed:

• ruff check daser tests benchmarks
• python3 -m compileall -q on changed Python files
• python -m pytest -q tests/unit/test_benchmark_correctness.py
• git diff --check
• vulture --min-confidence 60 audit; remaining findings inspected as dynamic/framework/protocol false positives

N/A / not run:

• Connector pytest was not run in this environment because torch/vllm are unavailable.
• Integration and benchmark/e2e runs are N/A for this deletion-only cleanup.
• Full pre-commit run --all-files was not run; the commit hook ran formatting on staged files.

Checklist

• pre-commit run --all-files passes
• Type hints and docstrings on all new/modified functions
• New features and bug fixes include tests
• No regressions — existing tests still pass