Fix register allocation in the presence of singled-valued phi nodes

[bjorng]

With optimizations disabled, the register allocator in beam_ssa_pre_codegen could allocate different variables to the same register if single-valued phi nodes in a try/catch construct were present.

Closes #7248

[RobinMorisset]

Thanks!

[github-actions]

CT Test Results

       2 files     296 suites   12m 29s ⏱️
   777 tests    775 ✔️ 2 💤 0 ❌
4 913 runs  4 911 ✔️ 2 💤 0 ❌

Results for commit 3cb4455.

♻️ This comment has been updated with latest results.

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[RobinMorisset]

This PR makes the following erlang test cases:

f(_V0) ->
    {_V0 and _V0, [ok || <<_:(_V0)/integer-unsigned>> <= port_to_list(_V0 orelse _V0)]}.

and

f(_V1) when (_V1 xor _V1); {_V1} ->
    erlang:insert_element(_V1 orelse _V1, ok, ok),
    _V1.

and

f(_V1) ->
    case ok of
        _ ->
            not _V1,
            [ok || _ := <<_:(_V1)/bytes>> <- (_V1 andalso _V1)]
    end.

fail to compile with variations of the following error message:

Internal consistency check failed - please report this bug.
  Instruction: {swap,{y,1},{x,0}}
  Error:       {uninitialized_reg,{y,1}}:

I can provide more test cases if it would be useful for debugging.

[bjorng]

I have now pushed a different solution that I hope has no undesirable side effects.