feature(argus): JIRA Issues support

[k0machi]

This commit adds support for interfacing with Atlassian's JIRA service,
allowing argus users to attach, delete and look up JIRA issues for test
runs.

Fixes #847

[soyacz]

Comments as we discussed. I'll return to this review after the adjustments.

[k0machi]

Demo (styles are unfinished right now):

[k0machi]

Still a couple of things to iron out, mostly fixups for logic that filters GithubIssues as it needs dispatch logic for both now.

[soyacz]

looks pretty good.
Good opportunity to add unit tests.

[k0machi]

@soyacz Ready for review. I've added some tests, but they're disabled because test env doesn't have a way currently to interact with github or jira.

[k0machi]

Demo:

[soyacz]

Overall LGTM, take only for these comments AI made if are legit and apply fixes if needed:

• frontend/Github/Issues.svelte:126-136 – label2color calls sha1 that is never imported in the module script (the only import is in the
  instance script), so the component won’t compile/run once this helper is used (JiraIssue imports it) and the UI breaks before
  rendering issues.
• argus/backend/service/issue_service.py:9-12 & argus/backend/tests/conftest.py:171-177 – IssueService now eagerly instantiates
  JiraService (which requires JIRA_SERVER/JIRA_EMAIL/JIRA_TOKEN and a live Jira connection). Any code path or test that just needs
  GitHub issues will now fail at construction time when Jira isn’t configured/available, even though the feature is optional and tests
  are marked skipped. This makes existing GitHub-only flows and CI brittle.
• frontend/Github/Issues.svelte:207-219 & argus/backend/service/jira_service.py:102/64 – Jira statuses are lowercased when created (“to
  do”, “in progress”) but refreshed with original casing and spaces; the UI filter hardcodes "todo"/lowercase keys. After a refresh,
  Jira issues can disappear from the filter (state key mismatch) and “To Do” is never matched. Normalizing states (including spaces) is
  needed both at ingest and refresh.
• argus/backend/service/jira_service.py:42-75 – Initial refresh window is set to Jan 1, 2025 and search uses the default search_issues
  page (max 50). On first run this skips all historical issues (nothing updated after that future timestamp) and ongoing sync misses
  updates beyond the first page when more than 50 issues changed. The Jira sync effectively becomes a no-op for most deployments.
• frontend/Github/IssuesCopyModal.svelte:45-56 – Uses document.querySelector/DOM scraping to copy HTML. This bypasses Svelte bindings
  and breaks the “avoid global selectors” guideline; a simple bind:this on the table would keep it rune-friendly and testable.

[soyacz]

indeed there are no mocks for these, but we don't need to much.
E.g for Jira, we can create one with just 2 methods (search and issue) that will return something and attach it to self.jira. I suppose AI will be able to do that.
You can try it in this PR or create a followup task we'll take later (I also can contribute).

[k0machi]

We can do that later yes, I've decided to put this off to a followup.

[soyacz]

just remember to add a task

[k0machi]

• frontend/Github/Issues.svelte:126-136 – label2color calls sha1 that is never imported in the module script (the only import is in the
  instance script), so the component won’t compile/run once this helper is used (JiraIssue imports it) and the UI breaks before
  rendering issues.

Incorrect.

• argus/backend/service/issue_service.py:9-12 & argus/backend/tests/conftest.py:171-177 – IssueService now eagerly instantiates
  JiraService (which requires JIRA_SERVER/JIRA_EMAIL/JIRA_TOKEN and a live Jira connection). Any code path or test that just needs
  GitHub issues will now fail at construction time when Jira isn’t configured/available, even though the feature is optional and tests
  are marked skipped. This makes existing GitHub-only flows and CI brittle.

Kinda irrelevant right now as we don't have CI for issues that utilizes these tokens. You can also say the same about GithubService (before and now).

• frontend/Github/Issues.svelte:207-219 & argus/backend/service/jira_service.py:102/64 – Jira statuses are lowercased when created (“to
  do”, “in progress”) but refreshed with original casing and spaces; the UI filter hardcodes "todo"/lowercase keys. After a refresh,
  Jira issues can disappear from the filter (state key mismatch) and “To Do” is never matched. Normalizing states (including spaces) is
  needed both at ingest and refresh.

Good catch. Will fix.

• argus/backend/service/jira_service.py:42-75 – Initial refresh window is set to Jan 1, 2025 and search uses the default search_issues
  page (max 50). On first run this skips all historical issues (nothing updated after that future timestamp) and ongoing sync misses
  updates beyond the first page when more than 50 issues changed. The Jira sync effectively becomes a no-op for most deployments.

Irrelevant, the history started a month ago. I will double check the search size however.

• frontend/Github/IssuesCopyModal.svelte:45-56 – Uses document.querySelector/DOM scraping to copy HTML. This bypasses Svelte bindings
  and breaks the “avoid global selectors” guideline; a simple bind:this on the table would keep it rune-friendly and testable.

Not sure what it means by "rune-friendly", but yes, we should use bind:this here.

[soyacz]

LGTM