[llvm] [AMDGPU] Fix end() iterator dereference in SILowerSGPRSpills (PR #88828)

[Christudasan Devadasan via llvm-commits]

cdevadas wrote:

Yes, `findDebugLoc` is the right one.

A simple case like the following would be enough.
Leave the first BB empty and a spill at any successor block would suffice. 
The test name 'empty_spill_block' is misleading. May be, empty_prolog_block?
---
name:            empty_prolog_block
stack:
  - { id: 0, name: '', type: spill-slot, offset: 0, size: 32, alignment: 4,
      stack-id: sgpr-spill, callee-saved-register: '', callee-saved-restored: true,
      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
machineFunctionInfo:
  hasSpilledSGPRs: true
body:             |
  bb.0:
    liveins: $sgpr0
  bb.1:
    renamable $sgpr0 = SI_SPILL_S32_RESTORE %stack.0, implicit $exec, implicit $sgpr32
    S_ENDPGM 0



https://github.com/llvm/llvm-project/pull/88828