Feature/use dpnp specific parfor

[ZzEeKkAa]

Introduce dpjit/dpnp specific pass. Instead of patching numba function update all call tree to generate right call. This unblocked setting dpnp specific instructions.

Checklist:

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• If this PR is a work in progress, are you filing the PR as a draft?

Fixes: #1270
Fixes: #146

[diptorupd]

@ZzEeKkAa overall the approach is good. I suggest you explore one thing. Instead of completely duplicating the upstream ParforPass and using our version of the upstream pass, maybe we create a supplementary pass. Our pass runs after the upstream pass and does the extra analysis that is needed to identify as parfor nodes dpnp array expressions that are missed by the upstream pass. The benefit of doing so, if we are able to do it that is, will be lesser maintenance overhead on our part.

[diptorupd]

Our pass runs after the upstream pass and does the extra analysis that is needed to identify as parfor nodes dpnp array expressions that are missed by the upstream pass.

That is just doConvertDPNPPass(self).run(self.func_ir.blocks)

[coveralls]

coverage: 84.819% (-0.06%) from 84.874%
when pulling a0d78df on feature/use_dpnp_specific_parfor
into 778bde3 on main.