bpo-45041: Simplify sqlite3.Cursor.executescript()

[erlend-aasland]

• increase test coverage for executescript
• minimise number of times we save/restore thread state
• optimise SQLite C API usage
• simplify error handling

https://bugs.python.org/issue45041

[erlend-aasland]

This PR greatly simplifies the SQLite statement error path. The sqlite3_stmt * pointer is now kept entirely inside the Py_BEGIN_ALLOW_THREADS/Py_END_ALLOW_THREADS wrapper, so it is very easy to track it from preparation to finalisation.

[erlend-aasland]

Code coverage after this PR and #28019:

Filename                      Regions    Missed Regions     Cover   Functions  Missed Functions  Executed       Lines      Missed Lines     Cover
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
prepare_protocol.c                 20                17    15.00%           4                 3    25.00%          22                15    31.82%
util.c                             65                13    80.00%           3                 0   100.00%          79                15    81.01%
module.c                          321                52    83.80%          11                 1    90.91%         241                52    78.42%
row.c                             216                44    79.63%          13                 1    92.31%         158                19    87.97%
microprotocols.c                   77                 8    89.61%           3                 0   100.00%          98                15    84.69%
connection.c                     1114               119    89.32%          48                 0   100.00%        1278               146    88.58%
module.h                            3                 0   100.00%           3                 0   100.00%          10                 0   100.00%
cursor.c                          796               103    87.06%          23                 0   100.00%         789               105    86.69%
statement.c                       345                20    94.20%          10                 0   100.00%         403                31    92.31%
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
TOTAL                            2957               376    87.28%         118                 5    95.76%        3078               398    87.07%

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @erlend-aasland for commit 9db16c3 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[erlend-aasland]

The AMD64 Windows8.1 Non-Debug PR and AMD64 Arch Linux Asan Debug PR buildbot failures are unrelated to this PR.

Manually rerunning failed buildbots. See python/buildmaster-config#275