[CSL-2435] Use 'TextEncoder' to measure correct length of UTF-8 strings

[nikolaglumac]

JavaScript length method (with ES6) actually counts unicode punycodes of the UTF-16 encoding of the given string. Apart from messing up with surrogate pairs (counting for instance 💩 as two characters because encoded with a pair or punycodes), it also leads to unexpected results when characters have a different UTF-8 and UTF-16 encoding.

Since requests are made using UTF-8 encoding, the Content-Length of these requests should match the actual length when payloads are UTF-8 encoded.

This PR is a port of #839

Acceptance tests results:

Before the fix:

After the fix:

[nikolaglumac]

I can confirm that all our acceptance tests are passing!

[nikolaglumac]

My manual testing has also proven that the fix works! 🎉

[nikolaglumac]

@DominikGuzei @darko-mijic we need to get this PR merged as soon as possible so that we can resume the release process...

[DominikGuzei]

@nikolaglumac you forgot to push the acceptance tests implementation?

[nikolaglumac]

No @DominikGuzei - I have added them in the first commit: 521d30a
(perhaps this was not the best idea in the world)...

[DominikGuzei]

@nikolaglumac ah sorry … i just saw the feature file and thought there are some implementation steps missing - but you just reused existing functionality 👍

[DominikGuzei]

@nikolaglumac i will merge this now as it works as intended. Please feel free to create an issue on YouTrack with the proposal by @cleverca22

[nikolaglumac]

Yes that single test captures the issue (in the code without the fix)