feat(dora): per-project DORA isolation for Jira boards shared by several projects

[kesamba]

Summary

When several DevLake projects point at the same Jira board and each narrows it with its own ExtraJQL, the DORA metrics that depend on incidents (Change Failure Rate, MTTR / failed deployment recovery time) counted the whole board for every project, so one project's incidents leaked into the others' numbers.

This PR records which project collected each board issue and filters on it end to end. It is split into commits that each build and pass tests on their own, so it can be reviewed commit by commit:

1. feat(core): add project_name to board_issues: new project_name column, part of the primary key so an issue in two projects' slices can be linked once per project, with default ''. Migration 20260820. Existing e2e snapshots that verify board_issues gain the (empty) column.
2. feat(jira): record the collecting project on board issues: project_name on _tool_jira_board_issues (migration 20260710) and ProjectName on JiraApiParams. The issue collector and extractor stamp the pipeline's project. ProjectName is omitempty in the serialized raw-data params, so pipelines without a project keep exactly the same params (no re-collection and no orphaned raw data on upgrade).
3. feat(jira): convert only the current project's slice of a shared board: the issue convertor reads only the current project's tool rows and stamps the project on domain board_issues.
4. fix(jira): resync the board when its ExtraJQL changes: the rendered ExtraJQL is folded into the board filter's change-detection key, so changing ExtraJQL triggers a full resync and issues that no longer match get unlinked from the board.
5. feat(dora): scope incidents and deployment links to the project's slice: ConvertIssuesToIncidents and ConnectIncidentToDeployment filter on board_issues.project_name.
6. feat(grafana): filter DORA incident panels by project: the same filter in the incident queries of the DORA dashboards (mysql: DORA, dora-by-team, dora-debug, dora-details-change-failure-rate, dora-details-failed-deployment-recovery-time, dora-details-timeto-restore-service; postgresql: DORA, dora-by-team).

Backward compatibility: if a board has no per-project stamps at all (single-project setups, or data collected before this change), every query keeps the previous whole-board behavior. If a board does have stamps, unstamped or NULL rows are treated as an ambiguous leftover pool and excluded, instead of being attributed to every project.

Does this close any open issues?

N/A

Screenshots

N/A

Other Information

• New unit tests run the real production queries against an in-memory SQLite DB. This adds gorm.io/driver/sqlite as a direct dependency; it was already in go.sum.
  • plugins/jira/tasks/issue_convertor_project_filter_test.go
  • plugins/jira/tasks/issue_extractor_test.go (TestExtractIssuesStampsProjectName)
  • plugins/dora/tasks/incident_from_issue_generator_test.go
  • plugins/dora/tasks/incident_deploy_connector_test.go
• The dora queries quote the reserved table column per dialect (backticks on MySQL, double quotes otherwise), so they run on MySQL, Postgres and SQLite.
• The full upstream CI passed on the fork, including test-e2e on both MySQL and Postgres (all plugin e2e suites plus test/e2e/remote and test/e2e/services), golangci-lint, migration-script-lint and check-grafana-dashboards: [fork CI] feat(dora): per-project DORA isolation for shared Jira boards kesamba/devlake#1

🤖 Generated with Claude Code