Import disks from ESXi to XCP-ng using XVA format

[fbeauchamp]

DO NOT SQUASH, review by commits.

Description

Changes :

1. load base disk then load snapshots, instead of loading a synthetic vmdk built from the base disk + snapshot
2. use xva import endpoint instead of vhd endpoint to import base disks into XCP:

• read base disk once instead of twice, de facto building thin disk for free
• use less memory
• smaller blocks means less empty blocks transfered

3. batch read when handling a sesparse snapshot ( esxi 6.5+)
4. don't read parent block if the current block of the snapshot is already full

Advice :

• please intall a xo in your esxi for the migration : most small read are done from XO to the esxi ,and theses are very sensible to latency

results

• XO installed as a VM directly on the ESXI8
• XCP-ng in local network over a 1Gbps link ( so 116MB/s max)
• speed limit of reading a flat.vmdk without doing anything ~330MB/s
• 1 VM , stopped, with 1 disk, 1 snapshot
• import duration on this branch : 4m46
• import duration on master : 71m25

network speed on the XO during the base disk transfer (read from esxi, write to xcp-ng)

network speed on the XO during the import of a snapshot of 1 VM, with 1 disk, with 1 snapshot, (read from esxi, write to xcp-ng)

review by commits

Checklist

• Commit
  • Title follows commit conventions
  • Reference the relevant issue (Fixes #007, See xoa-support#42, See https://...)
  • If bug fix, add Introduced by
• Changelog
  • If visible by XOA users, add changelog entry
  • Update "Packages to release" in CHANGELOG.unreleased.md
• PR
  • If UI changes, add screenshots
  • If not finished or not tested, open as Draft

[fbeauchamp]

@julien-f I finish testing it, and it gave me the same raw as master
I will do one more rebase and it is ready to merge