Stephan Rudlof <sr at evolgo.de> pointed out that
unpacking SmallInteger maxVal and trying to put the bytes back together
with Integer>>byte1:byte2:byte3:byte4: gives you a LargePositiveInteger
instead of a SmallInteger, and provided a correction.

On looking at #byte1:byte2:byte3:byte4:, I see a documentation problem.
The comment doesn't explain whether it is supposed to construct a
*signed* 4-byte integer or an *unsigned* 4-byte integer.

Could I ask for a change to the comment as well, so that it reads

    "Answer an integer in the range 0 to 2**32-1,
     given its base-256 digits (byte4 byte3 byte2 byte1).
     The result is a SmallInteger if possible, otherwise a LargeInteger.
     Only the most significant byte needs to be checked to decide which."