fix: fix proc scanning on linux 32bits without LFS

The commit dfd6f7b brought large file detection in the autoconf file
to fix process scanning on linux in x86 architecture. This is however a
bit flaky, because:
- process scanning won't work on systems that do not have LFS, which
  does not make sense (although I don't really know how autoconf detects
  it or not).
- the fix does not work if the compilation is done without the use of
  autoconf, and `_FILE_OFFSET_BITS` is not defined. This is for example
  the case for yara-rust, where the process scanning does not work on
  linux 32bits for this reason.

This breaks process scanning because without it, the pread function
on the mem file will fail if provided with an offset bigger than
INT32_MAX, which will always happen for processes.

A better fix is to use the pread64 syscall, which always takes a 64-bits
value for the offset. This is the actual linux syscall, so it is
completely safe.

See for example the exact same fix on chromium:

https://groups.google.com/a/chromium.org/g/crashpad-dev/c/IJLo8M2AQs0/m/Edyyau2LAwAJ

libyara/proc/linux.c

@@ -249,7 +249,7 @@ YR_API const uint8_t* yr_process_fetch_memory_block_data(YR_MEMORY_BLOCK* block)
   // target process VM.
   if (fd == -1)
   {
-    if (pread(
+    if (pread64(
             proc_info->mem_fd,
             (void*) context->buffer,
             block->size,
@@ -265,7 +265,7 @@ YR_API const uint8_t* yr_process_fetch_memory_block_data(YR_MEMORY_BLOCK* block)
     {
       goto _exit;
     }
-    if (pread(
+    if (pread64(
             proc_info->pagemap_fd,
             pagemap,
             sizeof(uint64_t) * block->size / page_size,
@@ -284,7 +284,7 @@ YR_API const uint8_t* yr_process_fetch_memory_block_data(YR_MEMORY_BLOCK* block)
       // swap-backed and if it differs from our mapping.
       uint8_t buffer[page_size];
 
-      if (pread(
+      if (pread64(
               proc_info->mem_fd,
               buffer,
               page_size,