[llvm] [MC] Allocate MCFragment with a bump allocator (PR #96402)

[Alexis Engelke via llvm-commits]

aengelke wrote:

I had a brief look at this now. In some cases, I could reproduce a slight max-rss increase in some cases, but I couldn't track down an actual increase in heap memory usage (massif) -- in fact, it shows a slight reduction. My current hypothesis would be memory fragmentation in the glibc heap (small holes are no longer re-used for fragments). But I wasn't successful in tracking down the cause. I won't have time for further investigation over the next days.

https://github.com/llvm/llvm-project/pull/96402