fix: filter Keep-internal fields from Jira kwargs

[zekebawt]

Fixes #5534

Fixes regression introduced in #5607 where Keep-internal workflow fields were being passed to Jira API, causing errors.

Problem

The kwargs block added in #5607 filtered priority values but passed all remaining kwargs to Jira fields. This included Keep-internal fields that Jira rejects with 'cannot be set' errors.

Reported by @Eliesmbr in #5607 comments.

Solution

Added explicit filtering for Keep-internal fields:

• enrich_alert, enrich_incident
• dispose_on_new_alert
• audit_enabled
• if, name, condition, foreach, throttle, provider

These are now skipped while legitimate Jira fields continue to work.

Testing

• ✅ Priority filtering (original fix: Handle optional priority field in Jira issue creation #5607 feature) still works
• ✅ Custom fields via custom_fields parameter preserved
• ✅ Keep-internal fields excluded from Jira API calls
• ✅ Syntax validation passes

Regression Fixed

Resolves: 'Field '''enrich_alert''' cannot be set. It is not on the appropriate screen, or unknown.'

[vercel]

@zekebawt is attempting to deploy a commit to the KeepHQ Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

Target branch is not in the allowed branches list.

[shahargl]

lgtm

[shahargl]

@zekebawt would you mind fix the "test docs" step?

[Eliesmbr]

I think also some unit tests are failing because of the new change. Would you mind fixing these @zekebawt ? Currently all new PR pipelines will fail.

[Eliesmbr]

The 5 failing tests in tests/providers/jira_provider/test_jira_priority_fix.py are not caused by your code changes in this PR. They were introduced in #5607 but never ran in CI because GitHub requires workflow approval for first-time fork contributors the PR was merged before tests were triggered.

The tests use @responses.activate to mock HTTP calls. Unlike @patch("requests.post") (used by the existing Jira tests in tests/test_jira_provider.py), @responses.activate lets the real requests.post() → Session.request() code path execute and only intercepts at the HTTP adapter level.

This hits a monkey-patch in keep/api/api.py:96-105 that replaces Session.request with no_redirect_request. When api.py is loaded more than once in the test process, original_request ends up pointing to no_redirect_request itself instead of the real method causing infinite recursion:

keep/api/api.py:102: in no_redirect_request
    return original_request(self, method, url, **kwargs)
keep/api/api.py:102: in no_redirect_request
    return original_request(self, method, url, **kwargs)
...
RecursionError: maximum recursion depth exceeded in comparison

The tests should use @patch("requests.post") instead of @responses.activate, consistent with the existing Jira provider tests in tests/test_jira_provider.py.

Unfortunately i can not commit to zekebawt:fix/jira-kwargs-regression.