[llvm] [BOLT] Support computed goto and allow map addrs inside functions (PR #120267)

[Maksim Panchenko via llvm-commits]

maksfb wrote:

> @aaupov @maksfb would you be ok with removing the NO-PIE case from the `indirect-goto-relocs.test` while finding a fix for the failure?

Does the test fail regardless of the change in this (#120267) PR? Yes, it's reasonable to back it out.

https://github.com/llvm/llvm-project/pull/120267