[FIX] Preserve user baggage header when setting Session ID

[marco-saia-datadog]

What does this PR do?

When a RUM Session ID is available, we set it in the Baggage header of intercepted network requests when using trackResources=true.

We don't account for pre-existing values in the Baggage header, so we end up overwriting it.

This PR adds support for user-defined Baggage headers by using comma-separated list strings, following the W3C Baggage specification.

Requirements for W3 compliance

NOTE: All invalid keys or values are dropped with a warning

• Entry Format
  Each entry must contain at least one = separating key and value.

• Key Validation
  Keys must follow RFC 7230 §3.2.6 token grammar.
  Allowed characters: !#$%&'*+-.^_|~0-9A-Za-z

• Value Encoding
  Values are UTF-8 encoded.

  Characters outside the allowed baggage octet range (!, #–+, -–:, <–[, ]–~) are percent-encoded.

  Spaces, commas, semicolons, quotes, backslashes, and non-ASCII characters are encoded.

  Example:

   "Amélie"  => Am%C3%A9lie
   "us east" => us%20east
  

• Properties (Metadata)

  Entries may include properties separated by semi-colons:

   key=value;prop1=value1;prop2
  

  Property keys must also follow RFC 7230 token grammar.

  Property values are percent-encoded using the same rules as the main value.

• Whitespace Handling

  Leading and trailing whitespace around keys, values, and properties is trimmed.

  Internal whitespace (e.g. "foo = bar ; p1 = one") is normalized.

• Limits (per W3C Spec §3.3.2)

  The resulting header must contain:

  • ≤ 64 list-members, otherwise a warning is logged.

  • ≤ 8192 bytes total size, otherwise a warning is logged.

  Entries beyond these limits may be ignored by consumers. We print a warning on our side, without dropping the entry.

• Order Preservation

  Entries are formatted in the same order they appear in the original Set.

• Logging and Error Handling

  Invalid entries and properties are not propagated.

  Each failure is logged via InternalLog with SdkVerbosity.WARN.

Review checklist (to be filled by reviewers)

• Feature or bugfix MUST have appropriate tests
• Make sure you discussed the feature or bugfix with the maintaining team in an Issue
• Make sure each commit and the PR mention the Issue number (cf the CONTRIBUTING doc)
• If this PR is auto-generated, please make sure also to manually update the code related to the change

[marco-saia-datadog]

The encoding is custom since baggage headers support characters that would otherwise be escaped by functions like encodeURIComponent.

The logic is:

• Converts the character to an iterable utf8 buffer
• Converts each byte value to a hexadecimal string using toString(16)
• Pads it to always have two digits with padStart(2, '0')
• Uppercases it for canonical form.
• Prefixes it with %.

For example, if you encode "é":

Buffer.from('é', 'utf8') -> [195, 169] (0xC3, 0xA9)

result -> %C3%A9

[cdn34dd]

probably not super important, but I quickly tried you branch and added a test:

    it('test', () => {
        const entries = new Set([
            'traceId=abc123;sampled=true;debug',
            'test1= this is a test'
        ]);
        const result = formatBaggageHeader(entries);
        expect(result).toBe(
            'traceId=abc123;sampled=true;debug,test1=this%20is%20a%20test'
        );
    });

however I believe the correct result should be:
'traceId=abc123;sampled=true;debug,test1=%20this%20is%20a%20test'

[marco-saia-datadog]

Thanks for trying it out 🙏 However, the result is correct, as keys and values are trimmed on purpose.

In the W3 spec:

baggage-string         =  list-member 0*179( OWS "," OWS list-member )
list-member            =  key OWS "=" OWS value *( OWS ";" OWS property )
property               =  key OWS "=" OWS value

Where OWS stands for "optional whitespace".

[cdn34dd]

Look good.