[ORC] Update ELF debugger support plugin to use AllocActions

[weliveindetail]

The DebugObjectManagerPlugin implements debugger support for ELF platforms with the GDB JIT Interface. It emits a separate debug object allocation in addition to the LinkGraph's own allocation. This used to happen in the plugin's notifyEmitted() callback, i.e. after the LinkGraph's allocation was finalized. In the meantime, it had to block finalization of the corresponding materialization unit to make sure that the debugger can register the object before the code runs.

This patch switches the plugin to use an allocation action instead. We can remove the notifyEmitted() hook and implement all steps as JITLink passes:

1. We keep fixing up section load addresses in PostAllocationPasses
2. We now emit the debug object already in PreFixupPasses, i.e. before the LinkGraph's allocation is finalized. This is great, because it can run in parallel to the LinkGraph's link phase 3.
3. We await the finalization of the debug object in PostFixupPasses and run the GDB JIT Interface registration in an AllocAction when finalizing the LinkGraph's allocation.

Main benefits are:

• plugin is ported to the new AllocAction approach
• we save at least one RPC round-trip
• concurrent emission of the debug object reduces wall-time
• once this landed, we can drop EPCDebugObjectRegistrar

This PR implements the basic switch. I want to add some more improvements to the plugin in a follow-up (mostly NFC then).

[weliveindetail]

FYI: the Windows build failed with an unrelated error

2025-11-13T18:23:06.0717371Z [10620/11138] Linking CXX executable bin\mlir-capi-pass-test.exe
2025-11-13T18:23:06.0718013Z FAILED: [code=4294967295] bin/mlir-capi-pass-test.exe 
2025-11-13T18:23:06.0728799Z lld-link: error: undefined symbol: void __cdecl mlir::ptr::registerConvertPtrToLLVMInterface(class mlir::DialectRegistry &)
2025-11-13T18:23:06.0730811Z >>> referenced by MLIRRegisterAllExtensions.lib(RegisterAllExtensions.cpp.obj):(void __cdecl mlir::registerAllExtensions(class mlir::DialectRegistry &))

[lhames]

If we switch to a dealloc action for deallocation I think that would simplify this a lot further: At that point DebugObjectManagerPlugin doesn't need to track the resources at all since they'll be covered by the actions attached to the memory allocation.

[weliveindetail]

Thanks for the feedback. Yes, exactly! I have a patch for it already: weliveindetail@2cf880f, but there are 2 open issues. I will sort it out into PRs next week. I think it's good to review in isolation.

[lhames]

Thanks for the feedback. Yes, exactly! I have a patch for it already: weliveindetail@2cf880f, but there are 2 open issues. I will sort it out into PRs next week. I think it's good to review in isolation.

LGTM. Thanks @weliveindetail -- I look forward to checking out the upcoming PRs too. :)