Enhance pattern tokenizer with multi-charset support

[ljzxc]

Support combined character sets within a single token

Changes

• Modified tokenize_format_string to handle combined character sets within a single pattern token
• Now treats patterns like {lu} as a single position that can use any character from the combined set
• Instead of creating separate tokens for each character class, concatenates all specified character sets into a single larger set
• For example, {lu} now creates one token containing all lowercase and uppercase letters, allowing for all combinations at that position
• Added better handling for empty strings and invalid charset specifiers
• Improved check for distinguishing between range patterns ({a-z}) and combined character sets

Impact

This enables more flexible pattern specifications for users who want to specify multiple character classes at a single position. For instance, {lud} will now generate all combinations with lowercase letters, uppercase letters, and digits at that position, rather than treating them as three separate positions in sequence.

Example

$ gorilla -p lol{lu}{lu} 
gorilla: (warning) missing mutation sets
gorilla: will generate 2704 words from a pattern lol{lu}{lu}
         sizes before mutations: 16224 bytes / 0 MB / 0 GB / 0 TB
lolaa
lolba
lolca
[...]

$ gorilla -p test{luds}
gorilla: (warning) missing mutation sets
gorilla: will generate 95 words from a pattern test{luds}
         sizes before mutations: 570 bytes / 0 MB / 0 GB / 0 TB
testa
testb
[...]
test}
test~
gorilla: finished in 322µs. 95 words -> 95 words

Ref #42

[andreiverse]

Looks good to me. Amazing work!

[ljzxc]

Hey, let me know if you need anything else to merge this. Thanks!