Follow up work to clean up the initial implementations for run and device calls

Creating an issue to track some items that are left as todos in PRs:

• In runtime/common/BaseRemoteRestQPU.h,: TODO - clean this up to not have to do this. Considering the module's map, or cudaq::details::getKernelName(). But make sure it works for
• In runtime/common/RuntimeMLIR.cpp: FIXME: don't use a recursive walk to DFS for FuncOps, which appear as children, in a Module.
• Add proper validation to not permit arbitrary/non-profile compliant calls in lib/Optimizer/CodeGen/VerifyQIRProfile.cpp
• reexamine filterSpecificCodePatterns in runtime/common/RuntimeMLIRCommonImpl.h
• get rid of --enable-run
• eliminate the use of a range of different CUDAQ_ environment variables to manually configure the compiler (including #3223 )
• Address https://github.com/NVIDIA/cuda-quantum/pull/3223#issuecomment-3176933095

Start by reviewing the TODOs in runtime/common/BaseRemoteRestQPU.h, runtime/common/RuntimeMLIR.cpp, lib/Optimizer/CodeGen/VerifyQIRProfile.cpp, and runtime/common/RuntimeMLIRCommonImpl.h, then read the discussion on PR #3223. The scope spans validation, runtime cleanup, --enable-run, and CUDAQ_ configuration; done means the listed follow-up items and the referenced comment are addressed.