This use of LR previously allowed overwriting it with arbitrary addresses
that aren't signed. Change this so we never return to an arbitrary LR.

Instead of loading the InterpreterTrampolineEntry address into LR directly,
use an ADR instruction to place into LR the address of a piece of code
that jumps to the InterpreterTrampolineEntry instead. This makes a difference
because BR is also constrained by BTI, whereas RET isn't.

An alternative would have been to `Call` instead of `Jump` to the target
bytecode and avoid the ADR instruction altogether, but I wanted to keep the
same behaviour with respect to the return stack that the existing code
exhibits.

Also add a comment to src/regexp/arm64/regexp-macro-assembler-arm64.cc for
a similar use of LR that should eventually be removed.

Bug: v8:10026
Change-Id: I24a13481f3fa416247dab8f9e5ae6f52f6b2ad42
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2764761Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Georgia Kouveli <georgia.kouveli@arm.com>
Cr-Commit-Position: refs/heads/master@{#73535}