fix(codegen/regex): allow vec growth on parse

[tristan-f-r]

Fixes #389. This is less efficient, but it's on codegen.

[jeertmans]

Hello @LeoDog896, thank you for your contribution!

Can you motivate why we need to have a possibly growing vector instead?
I see where the error might come from, but that would be great to document / explain it a little better, as it goes against the previous code comments left by the author of Logos.

[tristan-f-r]

Yep! I'll write this in a comment soon, but the original vec allocation never accounted for the size of the literals - it always assumed every ropable element would only have one item, and each item would be four Unicode codepoints long; since literals can be longer than four Unicode points, if there was more than eight tokens (two being the minimum size to create a concat literal), it would cause that integer overflow error.

The only way to get that size would be to walk through all the elements in the concat node and add up the lengths of all of these elements.

[jeertmans]

Sorry for the late review @LeoDog896, I was on holidays. This looks great and clear to me.

Before I merge this, did you check that long_concat_389 fails without your fix?

[tristan-f-r]

Sorry for the late review @LeoDog896, I was on holidays. This looks great and clear to me.

Before I merge this, did you check that long_concat_389 fails without your fix?

On current master, with only long_concat_389 added, it fails 👍

[jeertmans]

Looks good, thanks!