Mongoframes

Thanks to tylerganter for the pull request to significantly improve the performance of update_many.

In this release:

- `update_many` now uses bulk writes which significantly improves the performance of calling update_many.
- `unset` and `unset_methods` have now been added to the `Frame` class.
- A fix has been applied to the `ElemMatch` function which previously would raise an error if a raw condition was given instead of a `Condition` instance.

It's now possible to set the collection options for a frame within a given context, such as:

python

with MyFrame.with_options(read_preference=ReadPreference.SECONDARY):
my_frame = MyFrame.one()



This release also sees a move to pymongo>=3.9.0 requirements.

This release has moved the minimum pymongo requirement to 3.8+. To support this change we've replaced a number of pymongo deprecated method calls, including `update` with `update_one` and `count` with `count_documents` or `estimated_document_count`. This does not change the` update` and `count` methods / API for MongoFrames.

In addition I fixed an issue where using `$slice` in a reference projection wasn't correctly, for example in the expression:

python
projection={
'name': True,
'show_date_str': True,
'leads': {
'$ref': User,
'$slice': 1,
'first_name': True,
'last_name': True
}
}


This now correctly selects the first Id in the leads field to use as a reference when selecting the assoicated users