Add rule about Roman numerals being valid and compact

[suchow]

[mpacer]

However, this rule should probably allow for the more common variations in roman numerals, such as the use of Ↄ (in the context of, e.g., CIↃ = M = 1000 or IↃↃↃ = 50000).

However, though attested examples such as IIIX and VIX to represent 13 and 16 ("to conserve…the Latin terms tertio decimo and *sextodecimo"; [Cajori, History of Mathematical Notations Vol 1. p31 ¶ 48 ]) we should probably include them as invalid for modern purposes.

Open question should we allow the consistent-with-convention but unconventional construction XXCIII instead of LXXXIII in cases where a shorter word-form could find use (e.g., "Karpinski points out that the subtractive principle is found on some early tombstones and on a signboard of 130 B.C., where at the crowded end of a line 83 is written as XXCIII, instead of LXXXIII", ibid. p31 ¶ 48).