The compiler in ecl version 10.7.1 generates erroneous code for the negation
of bits in bit-stream-read-bits if bits is declared as an (unsigned-byte 8).
Hence we declare bits as a (signed-byte 8) for ecl, given that ecl does not
generate optimal code in either case, but correct code for the later case.
Whereas asdf system names are strings and hence case-sensitive, many users
of asdf rely on symbol names being treated specially and down-cased to
system names. That does not work with "Deflate" but does work with "deflate".
NB that this is a non-compatible change for existing users.
