feat: optimize memory usage and add --auto-bf option

[cyw-cc]

Apply three sets of optimizations:

1. Memory/threading optimizations (kalloc, kthread, Hash_Table, POA):

   • Improve memory allocation in kalloc.cpp
   • Optimize threading in kthread.cpp
   • Refactor Hash_Table.cpp for better memory handling
   • Minor POA.cpp adjustments

2. Levenshtein distance optimizations (Correct, Levenshtein_distance.h):

   • Add SIMD-accelerated banded BPM functions (AVX2/SSE)
   • Add Reserve_Banded_BPM_8_AVX2_only and Reserve_Banded_BPM_4_SSE_only
   • Expand Levenshtein_distance.h with optimized edit distance implementation

3. Automatic bloom filter detection (--auto-bf, id=366):

   • sys.cpp: yak_available_mem_gb() reads MemAvailable from /proc/meminfo
   • Process_Read.cpp: ha_scan_read_stats() samples read lengths (avg/N50/min/max)
   • htab.cpp: ha_auto_bf_shift() two-level decision logic
     • Level 1: disable bloom filter if available memory insufficient
     • Level 2: disable bloom filter for HiFi-like data (avg>8kb, N50>8kb)
   • Assembly.cpp: call ha_auto_bf_shift() before ha_ft_gen()

Test results (SRR11606870, 4M HiFi reads, 16 threads):

• Peak RSS: 24.92 -> 16.92 GB (-32%)
• Real time: 1467 -> 1171 sec (-20.2%)
• Output: identical to baseline across all 5 GFA files