gh-156207: Fix curses Textbox.gather() for double-width characters

[fedonman]

gather() now reads each line with in_wchstr(), whose count is a cell count and which skips a double-width character's continuation cell, instead of reading one cell at a time and reporting that second cell as another copy of the character.

This fixes only the gather() symptom of gh-156207. The insert-mode corruption and the curses.error at the last column are untouched: both need _insert_printable_char() to know a character's display width, which is a larger change.

$ ./python -m test test_curses -u curses
Using random seed: 2577236069
0:00:00 load avg: 3.88 mem: 22.8 MiB Run 1 test sequentially in a single process
0:00:00 load avg: 3.88 mem: 22.8 MiB [1/1] test_curses
== Tests result: SUCCESS ==
Total tests: run=181
Result: SUCCESS

The duplication only exists on the main branch, where gather() started reading cells with in_wch(), so there is no NEWS entry.

[serhiy-storchaka]

This all is right, but using the new API allows us to go further. We no longer need move() + _end_of_line(y) to get the count here. We can simply read to the end of line and then strip redundant spaces (except one). Probably we can even use in_wstr().