feat: improve cookie chunk handling via base64url+length encoding

[hf]

Improves cookie chunk handling by introducing a new cookie encoding scheme that includes the length of the encoded Base64 value. It will prevent reconstructing data from stale cookies.

Due to bad uses of this package, some cookie chunks are not being properly deleted. Meaning that if a session was encoded in 3 chunks now suddenly goes down to 2 chunks, the last chunk is not being deleted. When it gets reconstructed, all the 3 chunks get concatenated and parsed. In some situations this leads to an invalid UTF-8 sequence (mainly because Base64 packs 6 bits into 8).

This PR addresses this by implementing a different Base64 encoding of the chunks. Instead of just splitting up a Base64 string into chunks, the first chunk will now contain the length of the string that follows. This will prevent a leftover chunk from being parsed as valid.

The encoding is as follows:

base64l-<length of base64 encoded string as base 36>-<base64 encoding>

The library now checks for these conditions and emits warnings to let the developer know that they have a bug in their integration.

[kangmingtay]

Due to bad uses of this package, some cookie chunks are not being properly deleted.

Do we know what causes the chunks to not be properly deleted? From a DX perspective, it seems rather confusing to have to deal with knowing whether to set the encoding to base64url or base64url+length