Rethink the special-case for Elemwise/Blockwise InplaceOptimizer for "large graphs"

[ricardoV94]

Description

As brought up in #1494 it's odd to have a hard-coded 500 number at which we stop trying to inplace scalar graphs. We could consider:

1. Remove this behavior and address underlying performance issues (if we can reproduce)
2. Make it a config variable
3. Make it more cleaver, not just scalar, but perhaps variable.size if that's statically known as well

[jessegrabowski]

I think it should be a config value

[ricardoV94]

I am thinking whether we need that in the first place