dd takes a really long time to allocate memory with large buffers (tested with 1G)

Trying to do a non-blocking read with a buffer size of 1G takes 772ms on rust dd, but 3ms on gnudd. That's a massive difference and makes fzf-tab feel very sluggish (see https://github.com/Aloxaf/fzf-tab/issues/549) just by switching from gnu coreutils to uutils.

% time gnudd bs=1G count=1 status=none iflag=nonblock if=/dev/tty of=/dev/null
gnudd: error reading '/dev/tty': Resource temporarily unavailable
gnudd bs=1G count=1 status=none iflag=nonblock if=/dev/tty of=/dev/null  0.00s user 0.00s system 82% cpu 0.003 total
% time dd bs=1G count=1 status=none iflag=nonblock if=/dev/tty of=/dev/null
dd: IO error: Would block
dd bs=1G count=1 status=none iflag=nonblock if=/dev/tty of=/dev/null  0.06s user 0.72s system 99% cpu 0.772 total

Start with the dd command entry point and reproduce the reported non-blocking read using a 1G buffer and the provided shell commands. Trace where the large buffer is allocated, compare the behavior with gnudd, and consider the issue done when the Rust implementation no longer incurs the reported allocation delay while preserving the command's non-blocking behavior.