RFE: more convenient disk reuse

The recent 19.2.1 point release made Ceph more susceptible to bootstrap faults if non-pristine disks were used:
- changes to the bootstrap tooling to accomodate crimson/seastore
- attempts at making OSDs more robust by writing metadata labels in several places across the disk

To mitigate we improved the disk wiping procedure and detection of non-pristine disks in #616 and #621

We had previously decided to not auto-wipe disks for safety, preventing users from overwriting in-use data by mistakenly adding incorrect disk devices. However we should consider making disk re-use more convenient in cases where we can infer user intent to re-use a disk. For instance currently we consider a disk non-pristine and ask for explicit wiping if we detect any existing bluestore labels. In this case we could auto-wipe a disk if we find bluestore labels with a fsid that is not equal to the current clusters fsid -- in this case it seems likely that a user intended to re-use the disk; similarly we could auto-wipe if a disk was being added that does not have a disk or partition signature, which in-use disks would have.

Start by reviewing the disk-wiping and non-pristine-disk detection changes in #616 and #621, then trace the current handling of Bluestore labels and disk or partition signatures. Define and test safe reuse behavior for labels from a different cluster FSID and for disks without a signature, while preserving explicit wiping for ambiguous in-use disks.

Written by the indexing model from the issue text.