Allow large strings in nvtext benchmarks

[davidwendt]

Description

Removes the 2GB limit check from the nvtext benchmarks and adjusts the parameters to be consistent across the benchmarks.
Also converts the subword-tokenizer to nvbench and removes the unused word_minhash.cpp source file.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[davidwendt]

/merge