Code clean up

[roman-kulish]

@hipertracker do you mind another PR?

Changes:

• using xxhash for string hashing instead of Golang internal runtime.memhash (internal functions may not be safe)
• using official collator for sorting instead of Tidwall's, this should also slightly improve sorting performance
• removed size hints for unique words dictionary: if you don't know how much data you need to store, don't preallocate the space 🤷‍♂️
• in the last commit I accidentally used Latin version of words extraction, in this PR I switched back to the proper Unicode one.

This version is a bit slower than the previous one (because of use of Unicode word extraction and no dictionary preallocation), but it is how I would do it, if I am not chasing extreme performance.

Could you also add RAM usage to your benchmark?

Cheers

[hipertracker]

What is the best way to check RAM usage of the running Golang script? MacOS has a tool Activity Monitor but it can't display anything running less than a second. Golang script with sorting showed up in this tool and used about 50MB.