LocalDateRange's day-before-max infinity rule is untested, and a minimum end counts as infinity #257

Problem

LocalDateRange's "day before the maximum stands for infinity" rule (#224, PR #238) has no test. Seeding end >= LocalDate.MaxIsoValue.PlusDays(-1) → end >= LocalDate.MaxIsoValue in LocalDateRange.EndIsInfinity passes every suite: unit, NodaTime, conventions, EF, EF-Noda and the full live integration suite. The live NodaTime test uses an inclusive infinity] literal. The same seed on DateRange is caught. Found by the final pre-release analysis on add665c (pass A).

Also in the same function (DateRange.cs:502, LocalDateRange.cs:568): an upper end at the minimum date counts as infinity. DateRange [0001-01-01,0001-01-01].Shift(5) and LocalDateRange [MinIso,MinIso].Shift(5) come back unchanged. In PostgreSQL that range is [-infinity,0001-01-02), whose upper bound is finite. A closed end's exclusive form is the day after it, so the end == MinValue arm is wrong.

Change

Drop the minimum from the end-is-infinity test (an end is infinite only at the day before the maximum, or the maximum, per the read rule), and pin both sides per infinity-backed type.

Done when

LocalDateRange.CreateFinite(x, LocalDate.MaxIsoValue.PlusDays(-1)).Shift(Period.FromDays(-1)) keeps its end. [Min,Min].Shift(5) moves for DateRange and LocalDateRange. Both fail with the respective seed.