csearch sometimes fails on character classes containing only the same letter in upper and lowercase about codesearch HOT 4 CLOSED

Actually, it looks like it fails when the two-letter-character class is the 
first item in the regex.

laptop$ csearch 'd[gG]r'
/[XXXXX]/userids.txt:dgryski

I did a bit of investigation last night and it looks like the problem is with 
match.go:stepByte().  If we need to fold for this particular instruction, we 
uppercase the character 'c', but it doesn't get reset back to the lowercase 
version  when we move on to processing the next state -- the character 'c' is 
still the modified uppercase version instead of the original lowercase version 
that's actually in the string.

I've attached a patch to match.go to fix this, and two test cases to 
regexp_test.go.

This issue was closed by revision 5cd8d184e954.

Issue 19 has been merged into this issue.