Still more updates to support Volume layer activities.

[leftwo]

This is a re-do, re-work of what was in: Make crutest know extent info for each sub volume. #1474

Two new structs to describe a Volume:

pub struct VolumeInfo {
    pub block_size: u64,
    pub volumes: Vec<SubVolumeInfo>,
}
pub struct SubVolumeInfo {
    pub blocks_per_extent: u64,
    pub extent_count: u32,
}

Added a new method to the Volume type that will gather extent info from each sub-volume and return a VolumeInfo struct that describes the whole volume.

In the BlockIO trait, replaced query_extent_size with query_extent_info. Objects that support this call will return the RegionExtentInfo type filled out. Things that do not will just return None.

Updated crutest's RegionInfo struct to have a VolumeInfo field.

Updated the fill_sparse and span crutest tests to perform their work on all sub-volumes.

print_region_description looks like this:

----------------------------------------------
random read with 4 KiB chunks (1 block)
region info:
  block size:                   4096 bytes
  sub_volume 0 blocks / extent: 100
  sub_volume 0 extent count:    15
  sub_volume 0 extent size:     400 KiB
  sub_volume 1 blocks / extent: 100
  sub_volume 1 extent count:    15
  sub_volume 1 extent size:     400 KiB
  total blocks:                 3000
  total size:                   11.7 MiB
  encryption:                   no
----------------------------------------------

Added fill-sparse test to crutest cli

Coming in a follow up PR I would like to rename RegionInfo in crutest to be DiskInfo. This would just be name changes, no functional changes. I decided to break that out to reduce review overhead.

[mkeeter]

This doesn't make sense to me; the sum of blocks-per-extent across all subvolumes isn't a meaningful value.

In a subvolume-flavored world, there is no single extent size (es), so we should just remove it. Extent count ec remains meaningful, but should be a function on VolumeInfo, e.g.

impl VolumeInfo {
    /// Returns the total number of extent files in this volume
    fn extent_count(&self) -> usize {
        self.subvolumes.iter().map(|s| s.extent_count).sum().unwrap_or(0)
    }
}

[leftwo]

Yeah, for this I'm going to remove the whole crutest perf test completely, so this was just to make it compile.
It's all going away.

[leftwo]

Should also fix:
#1477

Partial fix for this #1455, though we may want to re-think how jobs are returned and displayed at the volume level.

We can probably call this one fixed, as we are full volume now: #1454

[mkeeter]

Is this the correct error message? It's confusing to me, because this message is used for both write_log.is_empty() and ri_option.is_none() cases. Above, we use "Internal write count buffer empty" for the former.

(same question applies below in a bunch of places)

[leftwo]

It's not the correct message any longer, but even better, it's also not possible.
If we are at the place where we check for a write log, then it must exist, as we now have an <Option<RegionInfo>> instead of an "empty" one. So, if the RegionInfo exists, then we have a write log and we can avoid this whole check. I'll make that change in all the places where we check for it.