Remove name argument from compute() generic

[ianmcook]

The name argument of the compute() generic is a vestige from the old days before dbplyr was broken out as a separate package. It would be best to remove it if doing so will not break any reverse dependencies.

Some dplyr backend packages need a method to force evaluation of lazy operations without assigning an external name to the materialized result where it's stored in the backend system. An example of this is the arrow R package.

It seems like this change should not break anything, because dplyr SQL backend packages should all use dbplyr::compute.tbl_sql() (which redefines the name argument) or else define their own compute() methods.

[romainfrancois]

That seems reasonable although I'm not that familiar with compute(). I'll try to keep this in mind for the next round of revdep checks.