Raise the python-core post-merge timeout to 75 minutes

[ciaranra]

The post-merge python-core job (the full core Python suite; PRs run only the fast-path lockfile check per #331's critical-path design) has been timing out at its 45-minute limit on every dev push since 2026-07-07 ("Dep bump (#322)" was the first). GitHub reports job timeouts as "cancelled," so it never showed as a red failure — the last ~17% of the suite simply hasn't executed on dev since then.

Data points:

• Last pass: 2026-06-28, at 40m34s of the 45m budget (already ~90%).
• Every dev push since Dep bump #322 dies at 45m with pytest progressing normally, reaching ~83% (mid qec/test_from_guppy_dem.py) — implying the grown suite needs ~55m.
• Suite growth since the budget was set: docs-CI generated examples, Render signed PHIR registers as two's-complement bits instead of a sign prefix #339, Default HybridEngine to the Rust classical interpreter, with parity fixes and #345 engine fail-fast #347's expanded parity coverage (~3,200 differential cases), plus any slowdown from Dep bump #322's dependency bumps.
• All other jobs in the workflow have ample headroom (lint 13m/30m, slow-postmerge 33m/90m, smokes 2–16m) — only python-core is miscalibrated.

Change: timeout-minutes: 45 → 75, restoring roughly the original ~35% headroom.

Follow-up (not this PR): profile per-test durations to attribute how much #322 specifically added, and consider sharding the qec tree into a parallel job if the suite keeps growing toward the new budget.