Test gha gateway

[linxinan-yt]

Verification: https://github.com/youtube/cobalt_sandbox/actions/runs/28919810566/job/85805412561?pr=858
Bug: 483488213

[github-actions]

🤖 Gemini Suggested Commit Message

---

build: Support browser tests in GHA gateway

Update the on-device tests gateway client to process browser tests
similarly to unit tests. This change enables host dependency
packaging for Android browser tests and adds the ability to provide
explicit gtest filters via the command line.

Additionally, update Android test target configurations to use the
gtest type for browser tests and remove JUnit test targets that are
no longer required for this flow.

Bug: 483488213

---

💡 Pro Tips for a Better Commit Message:

1. Influence the Result: Want to change the output? You can write custom prompts or instructions directly in the Pull Request description. The model uses that text to generate the message.
2. Re-run the Generator: Post a comment with: /generate-commit-message

[gemini-code-assist]

Code Review

This pull request updates the test target configurations for Android, removing JUnit tests and changing the test type of cobalt_browsertests to gtest. It also updates the on-device tests gateway client to support browser tests and explicit gtest filters. The review feedback highlights critical issues with target-specific arguments accumulating across loop iterations due to shared list references. Additionally, the reviewer advises against hardcoding target-specific filters and test-type overrides in the Python client, recommending more maintainable, configuration-driven approaches instead.

[gemini-code-assist]

The test_args list is mutated inside the loop (via test_args.extend(...)) without being copied or re-initialized per target. This causes target-specific arguments (like test_attempts) to accumulate across iterations and results in all targets sharing the same list reference.\n\nCopying test_args at the start of the block ensures each target gets its own independent list of arguments.

    if test_type in ('unit_test', 'browser_test'):\n      test_args = list(test_args)\n      if not device_type or not device_pool:\n        raise ValueError('Dimensions not specified: device_type, device_pool')

[gemini-code-assist]

Similarly to the unit_test/browser_test block, test_args should be copied here to prevent target-specific arguments from accumulating across iterations and to avoid sharing the same list reference across all targets.

    elif test_type in ('e2e_test', 'yts_test', 'yts_wpt_test'):\n      test_args = list(test_args)\n      test_target = target_data['target']\n      test_attempts = target_data.get('test_attempts', '')

[gemini-code-assist]

Hardcoding the gtest filter for cobalt_browsertests directly in the Python client code bypasses the filter JSON file mechanism and reduces flexibility.\n\nInstead of hardcoding this in Python, consider defining this filter in a dedicated filter JSON file (e.g., cobalt_browsertests_filter.json under the filter directory) so that it can be managed consistently with other test targets.

[gemini-code-assist]

Hardcoding a target-name-based check ('cobalt_browsertests' in target_name) to override the test_type to 'browser_test' is a maintainability concern.\n\nThis hack is introduced because the test type for cobalt_browsertests was changed from "browsertest" to "gtest" in test_targets.json. If we instead keep a distinct test type (e.g., "browser_test") in the JSON configuration and map it appropriately in the runner, we can avoid hardcoding specific target names in the Python client.