Heap profiler with leak tracking

[rdettai]

Description
This PR adds a very lightweight, feature gated, heap profiling endpoint. It superseeds the previous attempt in #5724

Using the jemalloc native profiling is complicated, it's hard to asses its performance impact, and the resulting traces are hard to interpret.

This PR proposes an custom and simple heap profiler:

• written so that the profiler has almost no performance impact when it's not being used actively. Every allocation/deallocation simply checks for an atomic ENABLED flag with relaxed ordering and does no extra work if that flag is false.
• the profiler can be activated with a webservice. It builds allocation statistics by call site and prints a backtrace when they exceed their last highest by a given threshold.

When started, the profiling allocates ~30MB of symbols.

How was this PR tested?
Python scripts, see results below.

[rdettai]

I ran some tests both during indexing (ingest V2) and search (float aggregation). On a pretty system (10 indexes with 1 milion docs ingested to each index):

Without the feature flag:

• indexing duration: 96.81
• avg search duration: 0.0058

With the feature flag but without starting the profiling:

• indexing duration: 96.80
• avg search duration: 0.0060

With the profiling activated with min_alloc_size=64KiB

• indexing duration: 96.77
• avg search duration: 0.0059

With the profiling activated for search only and min_alloc_size=1KiB

• indexing duration: 96.80
• avg search duration: 0.014

With the profiling activeted for indexing and min_alloc_size=1KiB

• heap profiling stopped, memory_locations full