[STABLE ABI] Porting overdrive.cpp

[pearu]

This issue collects tasks that block porting overdrive.cpp to use torch stable ABI.

• implement accessor template as a torch::stable::Tensor template method
  Fix available: [STABLE ABI] Add accessor template method to torch::stable::Tensor pytorch#161967
• can we use at::parallel_for in torch stable ABI code?
• expose AT_DISPATCH_FLOATING_TYPES to stable ABI, currently one need to implement the dispatch logic using switch block. Not a blocker but would nice to have.
  For a workaround, see [STABLE ABI] Porting forced_align #4078

^ @NicolasHug @scotts @janeyx99