Correct three texts a tester reads: 8.20.2a's IIN check, 10.4.4 step 12's sequence, and the capture adapter option

[jadamcrain]

Three places printed something wrong or misleading. Nothing a test checks changes.

• 8.20.2a step 2 (3.1r1) printed "Verify that the IIN field does not contain bits IIN()": a Null Error Response that allows every error bit added an empty check. isNullErrorResponse now leaves that check out, and the two other places that spelled the same rule by hand (8.23's range-read decline and 3.1r1's static input decline) use it, with the same descriptions as before.
• 10.4.4 step 12 (3.1r1) said it reads the response "w/ sequence(0x00)", but it reads the Select's response at whatever sequence the Select used. It now says "Read a single fragment response", as the test's polls do.
• tcp.capture.adapter was described as the adapter's "IP address", which is what a tester sees in the file -write makes, but it must be given as -adapters prints it, with the leading / (docs/manual-testing.md says so). The description now says that, with an example, and the error for an unknown adapter says how to give it.

Checks

• The -describe output of every procedure, for both revisions, at every subset level, with the defaults, a device profile, and every boolean option on and off: the only differences are 8.20.2a's empty IIN line going away and 10.4.4 step 12's line.
• mvn test passes.