Pre-compute Coefficients for Common Languages in CharAugmenter

[LSanselme]

Issue description

Textnoisr uses a coefficient to take into account repetitions in consecutive letters in natural language.

As @felix-martel-prl said in #7 (review) :

The next step would be to pre-compute this coefficient for a range of common languages
CharAugmenter(language="en") is better than CharAugmenter(natural_language_swap_correction=1.052).

It could indeed enhance readability, and make the code more easily usable for non-English languages.

Suggested Implementation Steps:

• Identify a set of common languages for pre-computation.
• Implement a mechanism to store and retrieve pre-computed coefficients.
• Update the CharAugmenter module to use pre-computed coefficients when available.