[ANNOUNCE] PySQLite 0.5.0

[Gerhard Häring]

It was time to make another release of PySQLite, because lots of bugs were
fixed and several sensitive changes were made, which need exposure in real
life.

Given the nature of the changes, I'd recommend you test 0.5.0 well before
replacing 0.4.3.

I'm in the middle of rewriting the documentation, examples and the website, so
you can expect a 0.5.1 release within a few weeks.

Here's the list of changes since the last release:

- Removed all old docs and examples. Started writing new ones.

- In sqlite.connect(), the parameter 'client_encoding' was renamed to
   'encoding'. 'client_encoding' still works, but is deprecated.
- Removed the usage of the warnings module for optional DB-API extensions.
- Added the type constant sqlite.UNICODESTRING.

- Added missing copyright statements for port/strsep.c, which was imported from
   FreeBSD.

- John E. Barham helped with adding a binary type to PySQLite. To insert
   binary encoded data into a column of type BLOB, encode it using
   sqlite.Binary() first.

- Columns with the column type 'UNICODE' are now automatically converted to
   unicode strings.

- BIG CHANGE: Release the GIL while the SQLite engine does its work. The
   PyThreadState is stored in the connection object.

- Plugged a memory leak that appeared in user-defined aggregates.

- Added missing MANIFEST.in (fixes bug #833821).

- Fixed bug #842630: .commit() is now a no-op in autocommit mode, instead of
   raising an exception. This behaviour is mandated by the DB-API 2.0.

- Fixed bug #808952: The enhanced PgResultSet wrapper was not used for rows if
   you used the fetchmany() method.

- Make user-defined aggregates actually work.

- Detect more column type names as "numeric" ones.