add x-stream-offset header to StreamReader responses

[kickster97]

WHAT is this pull request doing?

After stream offsets were removed from disk storage(#1479), the x-stream-offset header is added dynamically at read time. The shift?() path handles this correctly, but the read() path (used by StreamReader for the /api/queues/:vhost/:name/stream endpoint) was missed.

Fixes #1773

This PR adds an offset parameter to StreamMessageStore#read() and updates StreamReader#each to track and pass the offset through, so the header is added consistently across both code paths.

HOW can this pull request be tested?

Run the added apec

[claude]

No issues found.

The implementation correctly mirrors the existing offset-header pattern used in shift? and shift_requeued. Offset tracking across segment boundaries is handled properly (offset only incremented on successful read). The add_offset_header allocation is consistent with the existing consumer delivery path.

[claude]

Code review: No issues found.

The changes are minimal and correct:

• The offset value from find_offset was already being returned but discarded; now it is properly captured and passed through to read.
• The add_offset_header helper was already used in the shift? (consumer) path; this PR correctly extends it to the StreamReader path.
• No other callers of read exist, so the signature change is safe.
• No allocation concerns — StreamReader#each is not a hot path.

[snichme]

The read method is currently only used by the stream reader, I don't think it's a good design to add offset as an additional parameter to that method, it's somethign that can be handled outside the method.