coredump: implement v2 metadata format

this also picks up kcrash's v2 metadata files now.

v2 has a more refined workflow. instead of doing everything in kcrash's
file and then throwing it away, we now consume the kcrash file to begin
with and then create a rich metadata file as json based on the kcrash
and journal data.

v2 also uses the coredumpd-like naming pattern to get names that have no
collision risk.

the main motivation behind this is that we can then persistently keep
the files around on disk and build a more advanced reporting workflow on
top of this. notably drkonqi-coredump-gui can then show which crashes
have been reported, which may need reporting and what event_id they have
been reported as.

another use case this opens up is previous-boot reporting. e.g. when
kwin crashed on shutdown we can now more reliably detect it as not yet
seen by the user and throw up notifications on next login