Parallel notebook pre-execution (jcache local-parallel) for container repos

[mmcky]

#92 Phase 3: parallel notebook pre-execution for container repos — jcache project execute --executor local-parallel — compressing the cold case from 2,350s to ~950–1,550s.

• Pilot for OOM first: 16 GB runners, no worker knob (Decision 3 of Optimize preview builds: phased plan for the fastest PR previews (tracking) #92 — does 16 GB survive 4 concurrent kernels on the heaviest DP lectures?)
• Roll out to the weekly cache builds

This is also the first step of #2 (standalone execution check of built notebooks), whose runtime budget depends on it.