Pyappcache

Added

- A new FilesystemCache
- Added a new serialiser which uses Parquet instead of pickle for Pandas
dataframes: `pyappcache.serialisation.pandas.DataFrameAwareSerialiser`.
- Support for Python 3.12
- A new `BinaryFileSerialiser`

Changed

- Caches now work on buffers internally, rather than strs
- This is a breaking change for compressors and serialisers
- This seems to work much better for larger values (eg dataframes, csv files,
etc)
- SqliteCache will now use [incremental blob
I/O](https://www.sqlite.org/c3ref/blob.html) where possible (eg Python 3.11+)
- Fixed an issue with the default prefix being "pyappache"
- Sort out CacheControlProxy

Removed

- Support for Python 3.7

Added

- `prefix` can be passed as an arg to `Cache.__init__`
- Support up to Python 3.11

Changed

- Added a better repr for `SimpleStringKey`
- Fixed two issues in sqlite where expiry and eviction were not correct

Removed

- Support for Python 3.6

Added

- Improved documentation greatly (and put it on RTD)
- GPLv3 license

Changed

- Change the Key classes, a breaking change

Removed

Added
- Support for operating as a read-through/write-through cache
- `get_via` (update-on-read)
- `set_via` (update-on-write)
- Key "namespacing" (documentation to follow)

Changed
- pylibmc and redis are now optional dependencies
- eg install `pyappcache[memcache]` or `pyappcache[redis]` to require them
- Default cache prefix is now just "pyappcache" and slashes are added when building raw keys
- The required API surface of caches has been reduced further
- MemcacheCache will retry exactly once when pylibmc raises a ConnectionError
- in order to be robust against memcache restarts

Removed
- repr no longer defined on GenericStringKey

- First version