One decision for whether a procedure runs, so -disabled lists each once

[jadamcrain]

Changes

• One decision. Revision.skip(procedure, options, level) says why a procedure won't run on a device configured for a subset level: it isn't implemented, it doesn't apply at that level, or an option disables it. The runner's skip reasons, Revision.validate and -disabled all use it, along with -describe's "not applicable" message. Before, each decided on its own, and they had already drifted once (Validate only the procedures that apply at the run's subset level #64).
• -disabled lists each procedure once. A procedure that doesn't apply at the configured level and is also disabled by an option was listed twice: as not applicable and again with its options. For example, 3.1r1's level 1 read tests (8.13.2a, 8.14.2a, …) appeared twice at levels 2 and 3, and 8.26.3 twice at levels 1 and 2 with supports.assignNoClass = false. It is now listed once, as not applicable. The output is otherwise unchanged: the same groups, order and lines.
• The runner's skip messages and -describe's output are unchanged.

Checks

• -disabled output. Compared for both revisions at every subset level, with the defaults, a device profile, and every boolean option on and off, before and after. The only differences are the duplicate entries going away.
• -describe output. Unchanged.
• Tests. New tests check that -disabled lists each procedure once at every level and setting, and that the runner, the configuration check and -disabled agree on which procedures run. mvn test passes.