feat: add streaming download support for daily extract files

[be-ant]

Problem

get_daily_extract_file in core.py:526-534 calls requests.get() without stream=True. The wrap_return decorator's FILE_RESPONSE path (common.py:64-67) then returns response.content, loading the entire file into memory before returning it.

Impact

Daily extract ZIPs can be 50+ MB. In memory-constrained environments (AWS Lambda at 128–512 MB), this wastes memory and risks OOM errors — especially when multiple extracts are processed concurrently.

POC Evidence

A streaming download of a 53.77 MB daily extract file was benchmarked against the current buffered approach:

Approach	Peak Memory
Current (response.content)	~108 MB
Streaming (stream=True + iter_content)	~55 KB

Both requests (stream=True) and httpx streaming produced identical byte counts. The OFS endpoint includes a Content-Length header, so progress tracking is also possible.

Proposed Solution

Add a new method get_daily_extract_file_stream(date, filename, chunk_size=8192) that:

1. Calls requests.get(..., stream=True) (bypassing or extending the wrap_return decorator)
2. Returns an iterator (or the response object as a context manager) yielding chunks via iter_content(chunk_size)
3. May require a new response type (e.g., STREAM_RESPONSE) in the wrap_return decorator, or can bypass it entirely for this method

Example usage (desired API)

with ofsc.get_daily_extract_file_stream(date="2024-01-15", filename="extract.zip") as stream:
    with open("extract.zip", "wb") as f:
        for chunk in stream.iter_content(chunk_size=8192):
            f.write(chunk)

Bonus: additionalHeaders Bug

While reviewing the implementation, a bug was found in _base.py:220-223:

# line 221 merges headers correctly...
headers = {**self.headers, **additionalHeaders}
# ...but line 223 overwrites `headers` with self.headers, dropping the merge
headers = self.headers  # BUG: discards additionalHeaders

This means additionalHeaders passed by callers is silently ignored. Worth fixing alongside the streaming work.