sqlite improvements

[abonander]

• use direct blocking calls for SQLite in sqlx_macros
  • this also ensures the database is closed properly, cleaning up tempfiles
• don't send PRAGMA journal_mode unless set
  • this previously defaulted to WAL mode which is a permanent setting
    on databases which doesn't necessarily apply to all use-cases
  • changing into or out of WAL mode acquires an exclusive lock on the database
    that can't be waited on by sqlite3_busy_timeout()
  • for consistency, sqlx-cli commands that create databases will still
    create SQLite databases in WAL mode; added a flag to disable this.
• in general, don't send PRAGMAs unless different than default
  • we were sending a bunch of PRAGMAs with their default values just to enforce
    an execution order on them, but we can also do this by inserting empty slots
    for their keys into the IndexMap
• add error code to SqliteError printout
• document why u64 is not supported (this comes up from time to time so might as well fix it now)

@LovecraftianHorror I believe these changes are all backwards-compatible but I wouldn't mind some more eyes on it.

@ipetkov I can no longer repro #1929 with the changes from #1930 and this, although most of the credit goes to #1930

closes #1374

[ipetkov]

I can confirm that 4d49d66 fixes the original issue I had in my build 🎉

[penberg]

@abonander So how does one now set WAL mode for SQLite when using AnyPoolOptions? I could execute PRAGMA journal_mode=WAL; in after_connect() hook, but that seems bit clumsy.

[abonander]

If the database already exists it will remember WAL mode and sqlx db create will create the database in WAL mode by default.

To explicitly set WAL mode, see https://docs.rs/sqlx/latest/sqlx/sqlite/struct.SqliteConnectOptions.html#method.journal_mode

[penberg]

@abonander Right, but the problem I have is that I am using AnyPoolOptions, not the SQLite-specific API in sqlx. The after_connect thing seems to work, though.