[datlechin]

Summary

• tablepro_core::edit::temporal_parts finds the day, clock and BC era ranges in a date or time text with chrono's parse_and_remainder, after the commit's own year normalization (chrono_year, far_year_stand_in). It replaces the hand-written scanner in the Row Details clock, so the pickers and the commit read a value the same way.
• The ranges come from the text's shape and the values from its digits, so a part out of range (2024-02-30, 12:34:99, 24:00:00) keeps its place and a picker replaces it.
• The calendar finds the day through it too. A timestamp the commit reads but the calendar did not (12024-01-05 10:00:00 as PostgreSQL writes it, 2024-1-5 10:00 or 2024-01-05 10:00 kept as typed by SQLite) lost its time to 00:00:00; now only the day changes.
• A picked day drops a BC value's era, since GtkCalendar days are AD, and the calendar opens on today for a BC value. Text with a clock and no day gets the day before the clock, with the local offset where a zoned column needs one.
• The offset for a clock added on the first day chrono holds no longer overflows.

Test plan

• edit::temporal_parts tests: loose spellings, a year past 9999 and past chrono's range, the BC era, parts out of range, text the commit refuses.
• row_detail::tests::a_picked_day_keeps_the_time_of_any_timestamp_the_commit_reads fails on main; a_clock_on_the_first_day_chrono_holds_takes_an_offset panicked on main. Every existing clock and calendar spelling test passes unchanged.
• verify.sh test tablepro-core edit:: (46 passed), test tablepro row_detail (58 passed), lint.