fix(backfill): openIssuesTotal under-counts when open PRs are subtracted from GraphQL issues

[glorysr1209-png]

Summary

Closes #159. Count open issues via GraphQL issueTypes: ISSUE instead of subtracting open PR totals from repository.issues, which under-counts repos that have open pull requests.

Changes

• refreshRepoGithubTotals: issues(states: OPEN, filter: {issueTypes: ISSUE}).
• supplementOpenIssuesFromGraphQl: same filter + skip PullRequest nodes.
• Regression test: 10 open issues + 3 open PRs → openIssuesTotal stays 10.

Real Behavior Proof

• Unit test asserts totals are not reduced by open PR count when issues and PRs are reported separately by GraphQL.

What I ran

npm test -- test/unit/backfill.test.ts -t "openIssuesTotal"

What I observed

openIssuesTotal matches the issues-only GraphQL total; supplementUnderCountIfNeeded can still run when REST under-counts below the authoritative issue total.

[JSONbored]

@glorysr1209-png This is still blocked by TypeScript.

A few notes:

• Adding __typename and filtering PR-shaped nodes out of the GraphQL issue supplement is the right direction.
• The response type was not updated, so npm run typecheck fails on src/github/backfill.ts: Property '__typename' does not exist.
• Once the type is fixed, the test coverage should keep the issue-vs-PR distinction pinned.

Required changes:

• Add __typename?: string | null to the relevant GraphQL issue-node type.
• Keep the supplement path filtering pull requests before upserting issues.

Validation expected:

• npm run typecheck
• npm run test:coverage

[JSONbored]

@glorysr1209-png this is still blocked by typecheck.

A few notes:

• Counting issues without subtracting PR totals is the right problem to solve.
• Current validate fails because src/github/backfill.ts reads __typename from a type that does not declare it.
• The body has some proof, but CI needs to be green before this can move.

Required changes:

• Fix the GraphQL payload typing for __typename, or avoid reading it from that typed object.
• Keep the issues-only total regression test.

Validation expected:

• npm run typecheck
• Focused backfill tests
• Full validate pipeline

[JSONbored]

This is now stale/have been awaiting changes for a few days with no response. Please feel free to try again based on current codebase status if interested.