docs: calendar factory rows in What runs where — PostgreSQL's first date, the extremes with the conversions off, Quarter's error text #450

Found by a review of #191 (main PR #447 @ 879c82a, release/10.x PR #449 @ 684216b), checked on PostgreSQL 16.15 and 18.1. Docs only: the translation itself agrees with memory everywhere it was swept.

Problem

Three sentences about the calendar factories are imprecise.

1. PostgreSQL's first date. docs/efcore.md#what-runs-where (the LocalDateRange/YearMonthRange row), the EF Core README, the four calendar entries in test/oracles/translation-support.json and NodaCalendarFactoryClientSideTests' summary say "the years before 4713 BC have no PostgreSQL date". PostgreSQL's first date is 4714-11-24 BC:

   select '4714-11-24 BC'::date;   -- 4714-11-24 BC
   select '4714-11-23 BC'::date;   -- ERROR: date out of range
   

   The wording also leads with -infinity, which suggests the cell would be fine with Npgsql.DisableDateTimeInfinityConversions set. It isn't: the month and the year that hold PostgreSQL's first day can't be built either.

   select date_trunc('month', '4714-11-24 BC'::date::timestamp);   -- ERROR: timestamp out of range
   select date_trunc('year',  '4714-12-31 BC'::date::timestamp);   -- ERROR: timestamp out of range
   

   The decision to keep the NodaTime forms in memory stands. Only its stated reason is off.

2. The extremes. docs/efcore.md says "The first and last periods end at -infinity and infinity, as Npgsql writes DateOnly.MinValue and MaxValue". That holds only with Npgsql's infinity conversions on. With them off they are the days themselves, as the translator and CalendarFactoryExtremesTests already know. Also, the first period starts at -infinity; it doesn't end there.

3. The error text. "PostgreSQL raises an error that names the argument". Quarter(y, 5) and Quarter(y, 0) raise make_date's own error, which names the month the quarter became:

   select make_date(9999, 5*3-2, 1);   -- ERROR: date field value out of range: 9999-13-01
   select make_date(2026, 0*3-2, 1);   -- ERROR: date field value out of range: 2026--2-01
   

Change

• (1) Say "dates before 4714-11-24 BC, so the month and the year that hold PostgreSQL's first day have no start", and that this holds with the infinity conversions on or off. The same in all five places.
• (2) "The first period starts at -infinity and the last ends at infinity, as Npgsql writes DateOnly.MinValue and MaxValue (the days themselves with Npgsql.DisableDateTimeInfinityConversions)." Then update E76's quote.
• (3) Say the error names the argument or, for a quarter, the month it stands for. Or guard the quarter like the year, so its message names the quarter.

Test

Docs only. OracleCatalogueTests holds E76's quote to the reworded sentence.