BININT1 emits unsigned integer that fits into 1 byte.
BININT2 emits unsigned integer that fits into 2 bytes.

So they cover [0, 0xff] and [0, 0xffff] ranges including edges. However
we were emitting BININTX opcodes only for numbers being strictly inside
those ranges - i.e. (0, 0xff) and (0, 0xffff) - without edges. As the
result what could be emitted as "K\x00." was emitted as
"J\x00\x00\x00\x00." and so on.

Fix it by emitting BININT1 for [0, 0xff] inclusive and BININT for [0,
0xffff] also inclusive.