ggml-alloc: optimize free block shifting with memmove

[GermanAizek]

Replaced a manual for loop responsible for shifting elements in the chunk->free_blocks array with a single call to memmove. This change leverages an optimized standard library function for block memory operations, which can be significantly more efficient than a manual loop. memmove is designed to handle potentially overlapping memory regions correctly and is often implemented using highly optimized assembly instructions (like SIMD) or intrinsic functions, leading to improved performance during free block insertion in the memory allocator.

References:

• Performance of memmove vs. manual loops:
  https://stackoverflow.com/questions/11090176/is-it-faster-to-loop-and-copy-or-use-memmove
  (A discussion highlighting why memmove is generally faster due to compiler and library optimizations.)

• Inside memcpy and memmove Implementations:
  https://nullprogram.com/blog/2023/08/17/
  (Explores how memcpy and memmove are often implemented at a low level to achieve high performance.)

• C Programming Optimizations - General Techniques:
  https://www.geeksforgeeks.org/c-programming-optimizations/
  (Discusses various C optimization techniques, including the use of efficient standard library functions.)

• Memory Allocation Principles and Optimizations:
  https://www.cs.cornell.edu/courses/cs3410/2019sp/schedule/L24-MemoryAlloc.pdf
  (Lecture slides providing context on memory allocation strategies and where optimizations like this fit in.)

Co-Authored-By: Gemini 2.5 Pro (References and description commit changes)