Docassemble

Added
- The `pdf_concatenate()` function converts word processing files.
- The `flash()` JavaScript function.
Changed
- When using `complete_attribute`, list gathering will not
first seek a textual representation of the object.
Fixed
- Inclusion of images in RTF when a width not specified.
- `value()`, `defined()`, and `undefine()` no longer raise exceptions
when passed constants or internal Python names.
- Errors in list gathering.

Changed
- Google Drive synchronization is now handled as a background task so
that it will not be interrupted by a web browser timeout.
Fixed
- Uploads embedded in docx template files.

Added
- The `set_title()` function.
Changed
- Interview title functionality in the `metadata` initial block.
Fixed
- Image URLs to PDF page images with S3/Azure Blob Storage.

Fixed
- PDF page retrieval index error.
- File permissions issue in S3 that prevents server restart. S3 users
must upgrade the Docker image.
- Incorrect timestamps when downloading from persistent cloud storage.

Changed
- When using S3 or Azure Blob Storage, the files will now be cached
on the server file system and updated when they change.
- When using the Playground and S3 or Azure Blob Storage is in use,
and there are more than ten decorations, no preview images will be
shown in "Variables, etc."
- The values of `SavedFile` objects will be cached in memory for the
duration of a web request, so that when you create a new
`SavedFile`, attributes will be set from memory.
- The first argument to a `MachineLearner` object will no longer be
used as a literal `group_id`. Rather, an argument `'fruit'` it will
be converted into a group_id of the form
`docassemble.packagename:data/sources/ml-interviewname.json:fruit`.
You can still set a literal `group_id` using the keyword parameter
`group_id`.
- When an `initial_file` is given to a `MachineLearner` object, two
different formats can be used: an array, or a dictionary of arrays.
If a dictionary of arrays, the keys will be checked and if the key
matches the `group_id` (with the part before the final
colon, if any, removed), the array corresponding to that key will be
used.
Fixed
- Playground wizard not including pdf and docx template files in package.
- Playground wizard not including static files in package.
- Playground wizard pickle error when using S3 or Azure blob storage.
- Bug where key 'sought' not found.

Added
- The `pdf_concatenate()` function.
Fixed
- Missing dependencies for `RandomForestMachineLearner`.