crucible-downstairs verify continue in the face of errors

[leftwo]

If crucible-downstairs verify finds a context slot it can't read, it bails early:

BRM42220006 # /tmp/crucible-downstairs verify -d /pool/ext/cf5f8849-0c5a-475b-8683-6d17da88d1d1/crypt/zone/oxz_crucible_af8a8712-457c-4ea7-a8b6-aecb04761c1b/root/data/regions/67f34c5a-7205-49af-b440-3c0218ed2b19
Error: missing context slot for block 49360 in extent 18

Ideally, it could continue and scan the rest of the extents, but just mark the one it failed to read as failed when it gets to it.
This would help to verify a region that is open an active.

[leftwo]

Maybe the verify command should live in a stand-alone area, so it can better handle these situations that would otherwise be fatal to the actual downstairs.

[leftwo]

Maybe rather than re-use the exiting code's method of getting all context slots at the start, do it on an extent by extent basis, where we walk the extents and when you get to an extent, open it, read it, scan it, then close and move on. This might reduce the window of time between when we find the context slot for a block and when we scan that block, which would help in the active IO case. Just to reduce the false positives we see when scanning regions that are undergoing active IO.

It may be impossible to scan a region that is active without coordination with the running downstairs process, but we can probably reduce the window.