The default encoding in Python 3 is 'utf-8', which is not a good choice
for byte-level data pack/unpack because it uses multiple bytes for code
point >= 128.

This commit changes encoding being used to 'latin-1', which covers all
possible code points in range (0-255) with a simple one-to-one mapping.

This commit has fixed some *UnicodeDecodeError* in practice.

The `send_and_recv` function is very slow on receiving a large volume of
data because it only receives 2048 bytes at a time and keeps appending
these small chunks of data into a large array.

This commit sets the recv buffer to a larger value. This has provable
performance benefits: The running time of a Xlib program has been
reduced from 12 seconds to 0.5 seconds, a 24x speedup.