Fix memory corruption, use-after-free, and allocation bugs in set_value() and get_value()

[AlexJuca]

This PR fixes several severe memory-safety issues in our hash-table implementation.

Both set_value() and get_value() contained bugs capable of causing:

• heap corruption
• dangling pointers
• partial writes
• data loss
• crashes during normal operations

This PR rewrites the critical sections to ensure correctness, atomicity, and safe memory ownership.
Key Fixes

Improvements to set_value()

• Ensures old values are only freed after new allocations succeed
• Fixes incorrect cleanup for new vs existing entries
• Fixes bucket corruption on value-allocation failure
• Prevents double-free conditions
• Adds zero-length value support
• Makes update logic fully atomic

Improvements to get_value()

• Corrects malloc(sizeof(pointer)) → malloc(sizeof(value_entry_t))
• Performs a deep copy of value bytes
• Removes use-after-free where out->ptr was freed before return
• Ensures safe error handling on allocation failures
• Prevents callers from mutating internal storage

What is the outcome of this?

After this PR:

SET is memory-safe, atomic, and predictable.
GET returns stable, independent values.
No dangling pointers, no double frees, no overwriting internal state.

Instruments reports clean behavior for all tested GET/SET patterns.

Testing

• Manual GET/SET/UPDATE sequences
• Instruments profiling with intensive workload

Verified correct behavior on updates, overwrites, and missing keys

Next Steps (Optional Future PRs)

• Add read-only zero-copy get_value_borrowed() for performance
• Consider migrating hashtable to Robin Hood hashing
• Add unit tests specifically targeting memory safety