This patch adds direct code-gen support for a faster MOD intrinsic for
REAL types. Flang has maintained and keeps maintaining a high-precision
implementation of the MOD intrinsic as part of the Fortran runtime. With
the -ffast-real-mod flag, users can opt to avoid calling into the
Fortran runtime, but instead trigger code-gen that produces faster code
by avoiding the runtime call, at the expense of potentially risking bit
cancelation by having the compiler use the MOD formula a specified by
ISO Fortran.

This flags enables the compiler to generate most of the debug
information in a separate file which can be useful for executable size
and link times. Clang already supports this flag.
 
I have tried to follow the logic of the clang implementation where
possible. Some functions were moved where they could be used by both
clang and flang. The `addOtherOptions` was renamed to `addDebugOptions`
to better reflect its purpose.

Clang also set the `splitDebugFilename` field of the `DICompileUnit` in
the IR when this option is present. That part is currently missing from
this patch and will come in a follow-up PR.

This PR builds on the https://github.com/llvm/llvm-project/pull/158314
and adds the lowering support for `-gdwarf-N` flag. The changes to pass
the information to `AddDebugInfo` pass are mostly mechanical. The
`AddDebugInfo` pass adds `ModuleFlagsOp` in the module which gets
translated to correct llvm metadata during mlir->llvmir translation.

There is minor correction where the version is set to 0 in case no
-debug-version flag is provided. Previously it was set to 2 in this case
due to misreading of clang code.

This PR is a reapplication of
https://github.com/llvm/llvm-project/pull/142686

This reverts commit 895cda70a95529fd22aac05eee7c34f7624996af.
And fix attempt: 06f671e57a574ba1c5127038eff8e8773273790e.

Performance regressions and broken sanitizers, see #142686.

This PR adds the support for -gdwarf-N option where allows user to
choose the version of the dwarf. Currently N can be 2, 3, 4, or 5. This
is only the driver part of the change. Later PRs will propogate it to
the IR.

Fixes https://github.com/llvm/llvm-project/issues/112910.

This patch adds the flag -fexperimental-loop-fuse to the clang and flang
drivers. This is primarily useful for experiments as we envision to
enable the pass one day.

The options are based on the same principles and reason on which we have
`floop-interchange`.

---------

Co-authored-by: Madhur Amilkanthwar <madhura@nvidia.com>

From the CUDA Fortran programming guide: 
> If CUDA Fortran is enabled in compilation, either by specifying -⁠cuda
on the command line, and pre-processing is enabled by either the
-⁠Mpreprocess compiler option or by using capital letters in the
filename extension (.CUF, .F90, etc.) then the _CUDA macro is defined.

Move the definition of `_CUDA` to the compiler invocation.