Master issue: Improve ScanCode resources usage (CPU, RAM, Disk)

It would be great to optimize resources usage.
In particular, these are the likely key hot spots:
- memory as used in license detection (was typically 0.8 GB now closer to 1.5GB per process)
- memory when return large JSON scans (which are all re-loaded in memory assembled from the cache)
- CPU when running copyright detection
- CPU when running license detection

Some related issues:
- [ ] https://github.com/nexB/scancode-toolkit/issues/2725
- [ ] https://github.com/nexB/scancode-toolkit/issues/1950
- [ ] https://github.com/nexB/scancode-toolkit/issues/3711
- [ ] https://github.com/nexB/scancode-toolkit/issues/1894
- [ ] https://github.com/nexB/scancode-toolkit/issues/3593
- [ ] https://github.com/nexB/scancode-toolkit/issues/2547
- [ ] https://github.com/nexB/scancode-toolkit/issues/1014
- [ ] https://github.com/nexB/scancode-toolkit/issues/1809
- [ ] https://github.com/nexB/scancode-toolkit/issues/3143