Remove usage of project.documentation_type in tasks

[stsewd]

This depends on #4895, because we check if search/epub/pdf is supported by sphinx there

Ref #4638

[stsewd]

I'm passing only the doctype here rather than the whole config object, we only use the doctype, but if someone thinks it's useful passing the config I can change it.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@aace732). Click here to learn what that means.
The diff coverage is 90.47%.

@@            Coverage Diff            @@
##             master    #4896   +/-   ##
=========================================
  Coverage          ?   76.86%           
=========================================
  Files             ?      158           
  Lines             ?    10042           
  Branches          ?     1267           
=========================================
  Hits              ?     7719           
  Misses            ?     1983           
  Partials          ?      340

Impacted Files	Coverage Δ
readthedocs/projects/tasks.py	75.75% <90.47%> (ø)

[stsewd]

Looks like I touched an untested code, need to figure out where to add the tests

[stsewd]

Need to add more tests for the test god

[stsewd]

The last commits added the missing tests here

[agjohnson]

I think passing a dict of the validated config would be cleaner, but I don't have any strong opinions about it. If there are potentially any other cases where we'll need more access to the config dict in these tasks, lets just pass a dict object.

[stsewd]

@agjohnson this is the only place that I'm worried about passing a config object, we don't have one here.

[humitos]

We need to decide what to do with these projects: shouldn't we update this to say sphinx here and avoid this issues?

In [6]: Project.objects.filter(documentation_type='auto').count()
Out[6]: 514

We can also remove this chunk of code, https://github.com/rtfd/readthedocs.org/blob/4cf6a2ee8a86d6366e726ac53d696f3bae7dbdba/readthedocs/projects/models.py#L273-L277

Sentry issue: https://sentry.io/read-the-docs/readthedocs-org/issues/785965991/

[stsewd]

@humitos I can do another PR to solve that.

[ericholscher]

Seems reasonable. Feels like a lot of work to be sending the doctype around, where we don't use it much. I wonder if we can make the update_search function smarter so it can just figure out which type of docs they are, instead of passing it around a bunch before getting there.

[stsewd]

I wonder if we can make the update_search function smarter so it can just figure out which type of docs they are

Yeah, I'm getting more familiar with those bits of code here, but I guess we could kind of drop the usage of doctype to make some decisions here.

[ericholscher]

This should be updated once #5121 passes and it should then be much simpler I think 👍

[stsewd]

This comment looks outdated