Make use of posix_memalign for hfile buffer.

On AMD EPYC 7713 aligning to cache size boundaries makes a very
significant difference to fp->backend->read performance in the
kernel.  A modern Intel CPU did not demonstrate this difference.

x86 often have cache line size of 64 bytes, and apple Arm chips 128
bytes.  I haven't tested if Arm benefits from alignment during read
calls, but we can check size with sysconf(_SC_LEVEL1_DCACHE_LINESIZE).
However to avoid additional autoconfery I just picked 256 as it gives
us headroom and is simple.

Speed ups on the AMD EPYC:

time bash -c 'for i in `seq 1 30`;do cat < ~/lustre/enwik9| ./bgzip -l5 -@32 > /dev/null;done'

Unaligned
real    0m45.012s
user    10m7.661s
sys     0m58.770s

Aligned
real    0m30.717s
user    11m14.004s
sys     0m32.921s

It is likely this could improve other bits of code too.

configure.ac

@@ -326,7 +326,7 @@ HTS_HIDE_DYNAMIC_SYMBOLS
 
 dnl FIXME This pulls in dozens of standard header checks
 AC_FUNC_MMAP
-AC_CHECK_FUNCS([gmtime_r fsync drand48 srand48_deterministic getauxval elf_aux_info])
+AC_CHECK_FUNCS([gmtime_r fsync drand48 srand48_deterministic getauxval elf_aux_info posix_memalign])
 
 # Darwin has a dubious fdatasync() symbol, but no declaration in <unistd.h>
 AC_CHECK_DECL([fdatasync(int)], [AC_CHECK_FUNCS(fdatasync)])

hfile.c

@@ -113,8 +113,14 @@ hFILE *hfile_init(size_t struct_size, const char *mode, size_t capacity)
     // FIXME For now, clamp input buffer sizes so mpileup doesn't eat memory
     if (strchr(mode, 'r') && capacity > maxcap) capacity = maxcap;
 
+#ifdef HAVE_POSIX_MEMALIGN
+    fp->buffer = NULL;
+    if (posix_memalign((void **)&fp->buffer, 256, capacity) < 0)
+        goto error;
+#else
     fp->buffer = (char *) malloc(capacity);
     if (fp->buffer == NULL) goto error;
+#endif
 
     fp->begin = fp->end = fp->buffer;
     fp->limit = &fp->buffer[capacity];