Account for excessive read limits in macho parsing

[wagoodman]

Adds validation for size fields read from untrusted Mach-O binaries before allocating memory. A malicious binary with large DataSize, Count, or Length values could previously trigger memory exhaustion.

Changes:

• Added size limits for superblobs (50MB), blob counts (25), and individual blobs (16MB) -- these are fairly wide margins
• Validates that data ranges actually fit within the file before reading
• Caches file size to avoid repeated seeks during validation

Adding these checks did aggravate the linter, so a small refactor was done to decompose common parsing paths (e.g. parsing superblob and blob bytes).