Qualify exclusions by revision, so one configuration serves every revision #87

#86 checks each test.exclude.<id> against the revision being run, and an id the revision doesn't have is a configuration error. That breaks the one-configuration-for-every-revision design (#12): a configuration that excludes a 3.1r1 procedure can't run 2.7.

configuration has 1 invalid option(s):
  test.exclude.8.26.2 = Assign Class (function code 22) isn't supported yet (stepfunc/dnp3#480): revision 2.7 has no procedure 8.26.2

stepfunc/dnp3 runs both revisions in CI from one device.ini, and needs to exclude 3.1r1's 8.26.2.

Unqualified ids are also ambiguous across revisions, even where both have the id. 8.23 is Device Attributes in 2.7 and Specific Point Ranges in 3.1r1, so test.exclude.8.23.2.1 would exclude a different test depending on the revision.

Proposal

Qualify each exclusion with the revision it's for:

test.exclude.3.1r1.8.26.2 = Assign Class isn't supported yet (stepfunc/dnp3#480)

• A run applies only the exclusions for its own revision.
• Validation:
  • An unknown revision id is an error, as for -rev.
  • An id the named revision doesn't have is an error, whichever revision is running.
  • Exclusions for other revisions are otherwise ignored.
• Everything else from Exclude procedures through the configuration, reported as not run #86 stays as it is: the not-run result, the report, and -disabled.

To decide

• Whether to drop the unqualified form, or keep it for configurations used with a single revision. Dropping it avoids the ambiguity above.