TracingAllocator + http_init + http_cleanup + ASAN segfault

[rcoh]

with AWS_MEMTRACE_STACKS and the allocator provided by aws_mem_tracer_new, calling aws_http_library_init followed by aws_http_library_cleanup causes a SEGV:

==47935==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x00010759d723 bp 0x7000098c5710 sp 0x7000098c56d0 T4)
==47935==The signal is caused by a READ memory access.
==47935==Hint: address points to the zero page.
    #0 0x10759d723 in aws_hash_table_find hash_table.c:413
    #1 0x1075a04d9 in s_alloc_tracer_untrack memtrace.c:190
    #2 0x1075a0195 in s_trace_mem_release memtrace.c:441
    #3 0x10759d613 in aws_hash_table_clean_up hash_table.c:271
    #4 0x107594947 in aws_http_library_clean_up http.c:470
    #5 0x1066fa1f8 in aws_crt::http::clean_up::hee41a875d61c2824 mod.rs:23
    #6 0x10671fb58 in aws_crt::library_clean_up::h3e1713a99b77da82 lib.rs:171

This doesn't happen with AWS_MEMTRACE_NONE or AWS_MEMTRACE_BYTES

[jmklix]

Fixed with this PR: d09b75e