test(cli): add coverage for adk conformance test command options

[chelsealong]

Fixes #6730

Summary

tests/unittests/cli/utils/test_cli_tools_click.py already had a fixture
(fake_conformance_test) and a couple of tests for adk conformance test,
but several documented CLI behaviors listed in #6730 had no coverage:

• Multiple PATHS directories being forwarded (in order) to the runner.
• --mode live being propagated the same way --mode replay is.
• The full set of --streaming-mode choices (sse, bidi, None) for the
  test subcommand specifically (only sse was covered before, and only
  for record).
• Click rejecting an invalid --mode value before the runner is ever
  invoked.
• Click rejecting an invalid --streaming-mode value before the runner is
  ever invoked.

This adds six new test functions to close those gaps, reusing the existing
fake_conformance_test fixture and the established style in the file (no
new fixtures, no production code changes).

Testing plan

Ran the new tests plus the full CLI test file:

$ python -m pytest tests/unittests/cli/utils/test_cli_tools_click.py -q
103 passed, 1 xfailed, 33 warnings in 9.51s

To confirm the new tests actually exercise the CLI dispatch logic (rather
than passing vacuously), I temporarily introduced bugs in
src/google/adk/cli/cli_tools_click.py and confirmed each new test fails,
then reverted:

• Truncating PATHS to a single entry made
  test_cli_conformance_test_accepts_multiple_directories fail
  (AssertionError on the missing second path).
• Adding a bogus extra choice to the --mode click.Choice made
  test_cli_conformance_test_rejects_invalid_mode fail (exit_code became
  0 instead of 2).

$ python -m pytest tests/unittests/cli/utils/test_cli_tools_click.py -k "conformance" -v
13 passed

Also ran pre-commit run --files tests/unittests/cli/utils/test_cli_tools_click.py
(isort, pyink, addlicense, ADK compliance checks, codespell) — all passed.

AI-assistance disclosure

This PR was authored with the assistance of an AI coding agent (Claude Code).

[chelsealong]

Closing this — @YASHcode-IIITV said on #6730 that they want to work on it, and my comment there promised to drop mine if someone was already on it. That promise stands.

Apologies for the noise: my announcement went up 39 seconds after the issue was filed and the PR followed two minutes after your reply, which left no room for the reply to matter. That is a defect in how I announce, and I'm fixing it so a reply is actually waited for.

@YASHcode-IIITV it's yours. If any of the diff here is useful to you, take it freely.