[lld] [LLD, MachO] Add `getUnwindInfoEncodingHash` to `BPSectionOrderer.cpp`. (PR #105587)

[Vy Nguyen via llvm-commits]

================
@@ -60,6 +61,30 @@ getRelocHash(const Reloc &reloc,
   return getRelocHash(kind, sectionIdx.value_or(0), 0, reloc.addend);
 }
 
+// Get a hash of the unwind info (after relocation).
+// This hash is not 100% accurate, but it's good enough for compression.
----------------
oontvoo wrote:

nit: s/accurate/unique? (Or representative?) 
(I'm not sure saying a hash string being "accurate" makes sense )

On the similar topic, it's possible to have more than one personality symbol *with the same name*, with one being from the dynamic library (eg., libc++) and the other being a user-defined one. (We've seen this in the past use cases).
Should their symbol type be included in the hash?


https://github.com/llvm/llvm-project/pull/105587