Letter the tests 3.1r1 doesn't number

[jadamcrain]

New ids

3.1r1 leaves 103 tests unnumbered. Each one now gets its section's number followed by a letter for its place among the section's unnumbered tests, in document order:

Before	After
8.6.5-2	8.6.5b
8.14.2-3	8.14.2c
8.21-8.13.2-4	8.21d
8.22-8.14.2-16	8.22p

• No cross-references. An id no longer refers to another section. 8.21 and 8.22, which rerun 8.13's and 8.14's tests with double-bit inputs, are lettered in the same order as the tests they rerun.
• Can't be mistaken for 3.1r1's own numbers. 3.1r1 numbers only with digits, so 8.22a sits beside 8.22.1. There's no dot before the letter, so an id doesn't look like a step label such as TB2017-001's "1.a".
• Enough letters. The largest section needs 16 letters, a to p. A new test checks that each section's letters run a, b, c… with no gap.

New titles for the double-bit reruns

The 20 double-bit reruns now say which test each one runs. For example, "Double-bit Input Change, Qualifier 0x06 (as 8.14.2c)" replaces "Binary Input Change, Qualifier 0x06, with double-bit inputs".

Other mentions of the ids

Comments, test names, option descriptions, the README and docs/manual-testing.md all use the new ids, and their explanations of the scheme describe the letters.

Checks

I compared a dump of both revisions' trees, every procedure's -describe output, the disabled lists and Revision.validate, before and after, with the new ids mapped back to the old ones. It covered every subset level in three configurations. The only differences are the 20 new titles. mvn test passes.

Saved results

Results saved under the old ids don't count in a report under the new ones. No release has 3.1r1 yet, so this only affects local runs, which need running again. stepfunc/dnp3's conformance device.ini names some of these ids in a comment; that gets updated with its next dnp3c pin.