[Heap Trace Standalone] fix terrible Leaks perf on large records by using doubly linked list (IDFGH-8690)

[chipweinberger]

❌ This PR is rejected in favor or a new PR: #10478 ❌

This augments: #10127 (Edit: this PR will need to be rebased onto #10141)

Leaks: The original implementation memmove()'s the entire record buffer every time a record is freed. This is obviously terrible for perf.

My ESP32-S3 was not useable after ~300 records in SPI ram.

With this change, I can keep Heap Trace Standalone enabled without a noticeable impact on perf, tested up to 2540 records!

[chipweinberger]

Please wait on reviewing this PR.

• need to rebase
• plan to add a optional very simple hashmap lookup, for more speedup. It wont handle collisions. Just falls back to traversing the linked list if the record->address is not the first item it looked up.

typedef struct {
   heap_trace_record_t* record;
} heap_trace_hashmap_item_t;

// recommended size: 8x-12x the number of records
esp_err_t heap_trace_add_hashmap(heap_trace_hashmap_item_t* items, size_t count)

[chipweinberger]

Closing this PR in favor of a new refactored PR.

#10478