[Google] Vendor Transformer Adoption

[tanujkhattar]

Summarize the task
We now have cirq.CompilationTargetGateset and cirq.optimize_for_target_gateset abstractions and transformers to:

1. Specify a compilation target gateset that we want to compile a circuit to
2. Specify a bundle of transformers that should be executed in order to compile circuits to the target.

cirq-google still has the legacy cirq_google.optimized_for_sycamore method, which should be deprecated in favour of the above constructs. This work should be done as part of updating the device specifications in cirq_google and integrating compilation targets with devices, as proposed by @verult

Acceptance criteria - when is the task considered done?

Concrete action items should be:

• Figure out a way to construct compilation target gatesets from device specification protos and attach them to grid device metadata.
• The compilation targets constructed as part of 1 would contain the individual transformers executed in cg.optimized_for_sycamore method as gateset.postprocess_transformers
• We can then deprecate cg.optimized_for_sycamore(circuit, optimizer_type='sqrt_iswap') in favour of using cirq.optimize_for_target_gateset(circuit, gateset=sqrt_iswap_device.compilation_target_gateset). Also deprecate cg.optimized_for_xmon
• Update google specific tutorials like docs/google/best_practices.md to use these new constructs instead of cg.optimized_for_sycamore.

Part of #5028 and #5050