perf(notifications): respect server-recommended poll interval

[afonsojramos]

Fixes part of #3101

Problem

Polling uses only the user-configured fetchInterval and ignores the X-Poll-Interval header GitHub returns on the notifications list endpoint. The header advertises the minimum poll spacing the server currently wants (normally 60, raised when GitHub asks clients to back off), so Gitify keeps polling at the configured rate even when the server requests slower polling.

Change

• New pollInterval registry (src/renderer/utils/notifications/pollInterval.ts): forge clients report the latest server-recommended interval per account, and computeRefetchIntervalMs returns the user interval stretched to the slowest recommendation across the currently polled accounts.
• The GitHub client reports X-Poll-Interval from both the single-page and paginated notifications list fetches.
• useNotifications passes refetchInterval as a function, so TanStack Query re-evaluates the interval after every fetch and picks up server changes without a query-key change.

Forges that do not send the header never report, and polling falls back to the user interval unchanged. With the default 60s fetch interval and GitHub's normal X-Poll-Interval: 60 this is behavior-neutral; polling only slows when GitHub asks for more spacing (or when a user configures an interval below the server minimum).

Tests

• pollInterval.test.ts: stretch-to-slowest semantics, never faster than the user interval, ignoring intervals from accounts no longer polled, ignoring invalid values.
• client.test.ts: header capture on both the single-page and paginated fetch paths.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud