[Bug] Check mremap return value to prevent potential crash in MMapIO

[LHT129]

Description

Add return value check for mremap calls in MMapIO on Linux platform to prevent potential segmentation faults when memory remapping fails.

Background

In src/io/mmap_io.cpp, the Linux code path (non-Apple) has three mremap calls that do not check the return value:

1. WriteImpl - when extending memory mapping
2. ResizeImpl - when expanding memory mapping
3. ResizeImpl - when shrinking memory mapping

When mremap fails, it returns MAP_FAILED (i.e., (void*)-1). Directly assigning this to this->start_ causes:

• Subsequent memcpy(this->start_ + offset, data, size) writes to address -1 + offset, causing segfault
• All subsequent read/write operations on invalid memory addresses

The macOS branch (#ifdef __APPLE__) correctly checks for MAP_FAILED and throws an exception, but the Linux branch does not.

Requirements

• Add return value check for all three mremap calls
• Throw VsagException with meaningful error message on failure
• Align Linux error handling with macOS implementation

Technical Details

The fix follows the same pattern as the macOS branch:

void* new_addr = mremap(this->start_, old_size, new_size, MREMAP_MAYMOVE);
if (new_addr == MAP_FAILED) {
    throw VsagException(ErrorType::INTERNAL_ERROR,
                        fmt::format("mremap failed: {}", strerror(errno)));
}
this->start_ = static_cast<uint8_t*>(new_addr);

Acceptance Criteria

• All three mremap calls have return value checks
• Error handling matches macOS branch behavior
• Code compiles successfully
• Unit tests pass

Related

• Files modified: src/io/mmap_io.cpp
• Reference: mremap(2) man page - returns MAP_FAILED on failure

Notes

This is a defensive fix that prevents crashes in low-memory situations or when system limits are hit. The error message includes strerror(errno) for debugging.