fix(mcp): validate org fetch before committing session

[aashish00021]

Problem

An agent calling switch-organization with an organization id it cannot access is told the switch succeeded. The handler wrote orgId to the session cache and returned a success message before (and independent of) fetching the organization, so a failed fetch was silently ignored. Every later org-scoped call then failed with an opaque 403/404 that never mentioned the switch, and the stale orgId persisted in the token-keyed cache for its full 7-day TTL.

switch-project had the identical shape and was already fixed in #71976 — the fix was never mirrored onto switch-organization.

Closes #78629

Changes

• switch-organization now fetches the organization first; only on success does it commit orgId and the cached org to the session.
• On a failed fetch, it throws a descriptive error (naming organizations-get as the recovery path) instead of reporting success. The original API error is preserved as cause, matching switch-project's pattern for keeping recoverable not-found/no-access failures out of exception tracking.
• No change to the success-path response shape or to switch-project.

How did you test this code?

• Added services/mcp/tests/unit/switch-organization.test.ts, mirroring the existing switch-project.test.ts pattern: one case asserting a failed fetch does not commit orgId and preserves the error cause, one case asserting a successful fetch commits orgId and the cached org. Both pass (npx vitest run tests/unit/switch-organization.test.ts tests/unit/switch-project.test.ts — 6/6 passed, including the 4 pre-existing switch-project cases, confirming no regression there).
• npx tsgo --noEmit in services/mcp: no errors introduced by this change (confirmed identical pre-existing error count with and without the diff, via git stash/stash pop; all pre-existing errors are unrelated @posthog/quill UI-app build artifacts, not touching tools/organizations or tools/projects).
• oxlint --quiet and oxfmt --check on both changed files: clean.
• Did not manually exercise this against a live PostHog org/session — no live environment set up for this change; relying on the unit tests above plus the direct mirror of the already-shipped switch-project fix.

Automatic notifications

• Publish to changelog?

🤖 Agent context

Found via an isolated multi-repo scan for well-defined, unclaimed backend bugs. Root-caused by diffing organizations/setActive.ts against the already-fixed projects/setActive.ts, which shares the exact same fetch-then-cache shape and already documents the failure mode in its own comments. The fix, tests, lint/format, and typecheck were all done and verified in this session before opening this PR.

Autonomy: Human-driven (agent-assisted)

[trunk-io]

Merging to master in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here