Make template research possible with precomputed MSA

[Qrouger]

For now, create_individual_feature.py when using precomputed MSA flags just create .pkl and .json files for the MSA.
With this patch, the user can choose if he wants templates for the precomputed MSA or not with the flag --use_templates in the command create_individual_feature.py

[DimaMolod]

Hi @Qrouger and thanks for your PR!
I have two suggestions/questions:

1. the name of the flag is misleading and should be changed to reflect that --use-templates only works for mmseqs2 with precomputed a3m MSA files. This will not affect the default hmmer pipeline, which is not clear from the flag name/description. Maybe change to something like re-search-templates-mmseqs2 with detailed description?
2. Could you explain the motivation, why would users wanted to re-search templates? There is also an option to use custom templates in colabfold (custom_template_path), which seems to me as more common strategy.
   Thank you!

[Qrouger]

Hi @DimaMolod,
My motivation is mainly related to the availability of MSAs in the AlphaFold database. Pre-computed MSAs are now almost never accessible. I might be wrong, but I don’t know any easier or faster way to find templates than MMseqs. Using this could improve the accuracy of predictions when pre-computed MSAs are available.

I don't understand why you mention “re-search” templates instead of simply “search” for templates.

[DimaMolod]

I see, this is indeed a good reason :-) with a3m from AF DB and templates from mmseqs remote, there is often no need to create features at all. Something on top of our input features database
Could you still change the name of the flag to make it clear that it relates to mmseqs? Then I'll merge your PR.
Thank you!