Don't choke on lhmn_* in cleanup with range

[sj26]

When running an Lhm.cleanup with a range it tries to strptime every table name, but if there are unswitched new tables then it fails:

 > Lhm.cleanup(false)

 Existing LHM backup tables: lhmn_statement_lines.
 Existing LHM triggers: .
 Run Lhm.cleanup(true) to drop them all.

 > Lhm.cleanup(true, until: 1.week.ago)

 ...:in `eval': invalid strptime format - `lhma_%Y_%m_%d_%H_%M_%S' (ArgumentError)
    from .../lhm-2.2.0/lib/lhm.rb:64:in `block in cleanup'
    from .../lhm-2.2.0/lib/lhm.rb:63:in `select!'
    from .../lhm-2.2.0/lib/lhm.rb:63:in `cleanup'

This patch includes the table in cleanup over a time range. We can't really date this table, so can't apply the range logic, but it should just be duplicated data in the table so should be safe to delete.

[taonic]

+1 on the fix. But I'm wondering if we are introducing risk by removing lhmn_ all together when the intention of until is to remove lhma_ with the specified date range only.

[sj26]

Yeah I was wondering that too, @taoza, but this is generated duplicate data, not canonical data:

should just be duplicated data in the table so should be safe to delete

[tguo]

@sj26 Any chance the Lhm.cleanup is triggered while an online migration is in process and lhmn_ is being used to make copy?

[sj26]

@tguo sure, but if you're running your migrations in multiple places then I think you have bigger problems.

[sj26]

I'll close this one if there's no movement. :-)