Don't read all the files that we need to restore, just the collections in the beginning. The rest of them are read using Fetch on demand when they will be restored.
I haven't tested for any changes in memory consumption, but this brings down time taken for "Enumerating items in repository" to <5s even for huge number of files.
<!-- PR description-->
---
#### Does this PR need a docs update or release note?
- [ ] ✅ Yes, it's included
- [x] 🕐 Yes, but in a later PR
- [ ] ⛔ No
#### Type of change
<!--- Please check the type of change your PR introduces: --->
- [x] 🌻 Feature
- [ ] 🐛 Bugfix
- [ ] 🗺️ Documentation
- [ ] 🤖 Supportability/Tests
- [ ] 💻 CI/Deployment
- [ ] 🧹 Tech Debt/Cleanup
#### Issue(s)
<!-- Can reference multiple issues. Use one of the following "magic words" - "closes, fixes" to auto-close the Github issue. -->
* https://github.com/alcionai/corso/issues/3011
* closes https://github.com/alcionai/corso/issues/3440
* closes https://github.com/alcionai/corso/issues/3537
#### Test Plan
<!-- How will this be tested prior to merging.-->
- [x] 💪 Manual
- [ ] ⚡ Unit test
- [x] 💚 E2E
7d1d9295eb