temp workaround for false approximate match

[dataders]

fixes: #11

#8 might be closer to the permanent solution, but I'm still bottoming out on root cause and will likely need to meet w/ @jtcohen6 sometime next week.

this isn't a permanent fix, but it will work in the short-term, while we dive deep on the source of the adapter.get_relation issue

an overly-simplified normal operation for the table materialization is:

1. get the database, schema and identifier of the relation we are trying to make
2. go look and see if a relation with the same three-part name already exists and assign result to old_relation
   1. if yes, return an Relation object with the three-part name, and a type property (so we know if it exists as a view or as a table)
   2. if no, return None
3. if old_relation is not None, drop the pre-existing relation.

however, because we currently do not support views, if we assume that there are no views currently in the database, then we don't need to step 2 above. instead we can always just call DROP TABLE <TABLE_NAME> IF EXISTS, because:

1. the relation should always exist, and
2. if the relation doesn't already exist, the statement won't fail because of the ... IF EXISTS clause

[octavianzarzu30]

verified it works.

[dataders]

@octavianzarzu30 if this looks good to you, I can release this change as 0.21.3 tomorrow.

[octavianzarzu30]

yep looks good, have even tried it today :)

[ima-hima]

One formatting change, one documentation request.

[ima-hima]

lgtm.

And thanks for the additional comment. It actually did help my conception.