Binary Ninja uses too-low buffer size when reading bndb

Version and Platform (required):

• Binary Ninja Version: 4.3.6879-dev (333734f4)
• OS: macOS
• OS Version: 15.3 (24D60)
• CPU Architecture: Apple M3 Max

Bug Description:
One of my bndbs has grown to 11GB. I've noticed that when I open the bndb, Binary Ninja spends almost 30 seconds just calling pread to load the bndb into memory. During this time, the UI is frozen. For the pread calls, the offsets are not always sequential but often are, while the read size is always 1024 bytes.

1024 bytes seems like way too low a read size. Here is a simplistic comparison using dd (although this is sequential):

% dd if=main_301.bndb of=/dev/null bs=1024 
[..]
11609607168 bytes transferred in 13.495961 secs (860228269 bytes/sec)
% dd if=main_301.bndb of=/dev/null bs=1048576
[..]
11609607168 bytes transferred in 0.684243 secs (16967082116 bytes/sec)

(After this phase of loading is done, the UI becomes responsive again and Binary Ninja starts using CPU on multiple threads while the progress bar ticks up.)

Consider implementing some sort of cache so that you can use a larger read size… or just use mmap.

Also consider fixing the UI to not be unresponsive while reading data.

Start by tracing the bndb loading path and its 1024-byte pread calls; the issue names no source files or tests. Compare larger reads or mmap while checking how loading affects UI responsiveness. Done means large bndb files load without excessive tiny reads and the UI does not remain frozen during the read phase.