Metapredict

Changes:

* Merged pull request from FriedLabJHU to make f-strings more Pythonic. Thanks!!
* Changed `return_normalized` keyword to `normalized` in `meta.predict_pLDDT()` for consistency with other functions
* Added sanity check in case a passed sequence is an empty string (h/t Broder Schmidt)
* Added docs for the mode keyword in `meta.percent_disorder()`, so this is actually obvious to understand (h/t Broder Schmidt)
* Added several additional tests and updated the docs

Changes:

* Some minor bug fixes and updates to code

Changes:

* Merged pull request from FriedLabJHU to fix keyword name `cutoff` to `disorder_threshold` in `meta.percent_disorder()`. Thanks!!

* Added the `mode` keyword into `meta.percent_disorder()` so disorder can be predicted in terms of what percentage of residues fall within IDRs, as well as what percent are above some fixed threshold.

Changes:
Fixed bug in metapredict-name command that could result in the organism name being named twice in the title of the graph.

Changes:
Added functionality to graph the disorder of a protein by specifying its common name using the *metapredict-name* command.

Changes:
Massive update to the network behind metapredict to improve accuracy. Implementation of code to keep the original network accessible to users. Changes to predict_disorder_domain functions where a DisorderObject is no returned and access to values are used by calling properties from the generated object. Graphing functionality updated to accommodate new cutoff value for the new network at 0.5. If the original metapredict network is used, then the cutoff value automatically resets to the original value of 0.3. Tests updated. Added metapredict-predict-idrs command to the command line. Added ability to predict disorder domains from python using external scores.