Add eligible segment filter for bandits

[tmancey]

Resolves brave/brave-browser#13395

Submitter Checklist:

• There is a ticket for my issue.
• Used Github auto-closing keywords in the commit message.
• Wrote a good PR/commit description
• Added appropriate labels (QA/Yes or QA/No; release-notes/include or release-notes/exclude; OS/...) to the associated issue
• Checked the PR locally: npm run test -- brave_browser_tests, npm run test -- brave_unit_tests, npm run lint, npm run gn_check, npm run tslint
• Ran git rebase master (if needed).
• Requested a security/privacy review as needed.

Reviewer Checklist:

• New files have MPL-2.0 license header.
• Adequate test coverage exists to prevent regressions
• Major classes, functions and non-trivial code blocks are well-commented
• Changes in component dependencies are properly reflected in gn
• Code follows the style guide
• Test plan is specified in PR before merging

After-merge Checklist:

• The associated issue milestone is set to the smallest version that the
  changes has landed on.
• All relevant documentation has been updated, for instance:
  • https://github.com/brave/brave-browser/wiki/Deviations-from-Chromium-(features-we-disable-or-remove)
  • https://github.com/brave/brave-browser/wiki/Proxy-redirected-URLs
  • https://github.com/brave/brave-browser/wiki/Fingerprinting-Protections
  • https://github.com/brave/brave-browser/wiki/Brave%E2%80%99s-Use-of-Referral-Codes
  • https://github.com/brave/brave-browser/wiki/Custom-Headers
  • https://github.com/brave/brave-browser/wiki/Web-Compatibility-Exceptions-in-Brave
  • https://github.com/brave/brave-browser/wiki/QA-Guide
  • https://github.com/brave/brave-browser/wiki/P3A

Test Plan:

Run build with features cli args, e.g. --enable-logging=stderr --vmodule="*/bat-native-ads/*"=9,"*/brave_ads/*"=9,"*/brave_user_model/*"=9,"*/bat_ads/*"=9 --brave-ads-staging --rewards=staging=true --disable-features="TextClassification" --enable-features="EpsilonGreedyBandit"

Case 1: Default models enabled (Text Classification and Purchase Intent)

• Verify that Text Classification and Purchase Intent models are enabled and Epsilon Greedy Bandit is disabled: Text classification feature is enabled and Purchase intent feature is enabled
• Trigger ad notification and verify that ads are served as usual via text classification
• Process purchase intent segment and verify keywords are extracted:
  Extracting purchase intent signal from visited URL and Extracted purchase intent signal from visited URL
• Verify that bandit is processing (despite the model being disabled, that is by design)
• Verify that the model is not suggesting segments to the ad serve pipeline. Shouldn't see Exploring epsilon greedy bandit segments: ... or Exploiting epsilon greedy bandit segments: ...

Case 2: Variant models enabled (Epsilon Greedy Bandit and Purchase Intent: --disable-features="TextClassification" --enable-features="EpsilonGreedyBandit")

• Verify that Epsilon Greedy Bandit and Purchase Intent models are enabled and Text Classification is disabled: Epsilon greedy bandit feature is enabled and Purchase intent feature is enabled and Text classification feature is disabled
• Process purchase intent segment and verify keywords are extracted:
  Extracting purchase intent signal from visited URL and Extracted purchase intent signal from visited URL
• Verify that bandit is processing
• Verify that the model is suggesting segments to the ad serve pipeline Exploring epsilon greedy bandit segments: ... or Exploiting epsilon greedy bandit segments: ...

Case 3: Restart after one round of training and make sure state wasn't deleted in user Preferences JSON key epsilon_greedy_bandit_arms

[moritzhaller]

LGTM++

[moritzhaller]

CI failing on macos due to known intermittent rollup issues