Should john smith, v read like JOHN SMITH, V under lenient_comma_suffixes=False? #727

T4 (the bare letter as an initial, #723) says a name written wholly in one case reads as its capitals. With the default policy that holds. Under Policy(lenient_comma_suffixes=False) it does not, for the one-letter roman numerals:

>>> P = Parser(policy=Policy(lenient_comma_suffixes=False))
>>> P.parse("john smith, v")     # given 'john', family 'smith', suffix 'v'
>>> P.parse("JOHN SMITH, V")     # given 'V', family 'JOHN SMITH'
>>> P.parse("smith, john, i")    # suffix 'i', no report
>>> P.parse("SMITH, JOHN, I")    # suffix 'I', reports comma-structure

The likely cause is that the strict comma-suffix test vetoes a one-letter numeral by its shape as written, a bare capital looking like an initial. It doesn't read the initial tag, which both spellings now carry. This predates #723 and also fails on 16ee285. The design-docs review of PR #726 found it, measuring 4 disagreeing shapes for i/v and none for other letters. It is recorded in decisions/T4.

To settle before implementing:

• Which reading is right? Should the strict test read the tag, so the lowercase spelling reads as the capitals do (john smith, v → given v)? Or does the knob's documented meaning, a strict v1-style token test, make the written shape the point?
• Coverage: test_a_one_case_name_reads_as_its_capitals runs the default policy only. Should its grid add the strict knob, and are there other policy knobs it should run under?