[clean strict optional] Fix another 32 errors

[ilevkivskyi]

This cleans --strict-optional for:

• treetransform.py
• typeanal.py
• maptype.py
• errors.py

There are still 2 errors in the last one, but one of them is a typeshed issue (will report it tomorrow), and another should be fixed when @elazarg will finish #1855.

There is something interesting that I have noticed: there are around 150 errors only because Instance.type has type Optional[TypeInfo]. However, it could be None only right after de-serialization. It seems that these 150 errors could be fixed either by tweaking de-serialization (use a temporary class NotReadyInstance?) or just add "semi-automatically" some 150 asserts.

[JukkaL]

I've noticed the Instance.type issue myself, and it would be better to have a dummy TypeInfo than None. With this change, if the dummy value is accidentally used somewhere, it will be obvious where the value is from, whereas a None value could come from anywhere. One way to do this would be to have the dummy value raise an exception on any attribute access.

[JukkaL]

Thanks for helping us move towards strict optional! Feel to merge as is, or fix the Instance.type issue first.

[ilevkivskyi]

I think it will be easier to merge this now and make a dedicated PR for Instance.type.