fix(huggingface): fixing aiohttp not working with VCR if passing filename

[tito]

My current analysis of the bug of aiohttp not working under VCR on some tests can is this. Here is the breakdown using huggingface object detection

1. object_detection accept str/Path/bytes/BinaryIO (?)
2. When passing str, actually huggingface internally open and pass to aiohttp a io.BufferedIO() of the file at str.
3. When VCR is used, their implementation prior sending the request to aiohttp, will create their own Request
4. When building the request, the body is read()
5. Later, when the request is done by aiohttp, as it's a file read() is called, but there is nothing to read anymore.

Here is a fix before trying to fix VCR.py

[coveralls]

Pull Request Test Coverage Report for Build 12840122239

Details

• 3 of 3 (100.0%) changed or added relevant lines in 2 files are covered.
• 45 unchanged lines in 9 files lost coverage.
• Overall coverage increased (+15.3%) to 95.824%

---

Files with Coverage Reduction	New Missed Lines	%
scope3ai/api/tracer.py	1	92.31%
scope3ai/tracers/huggingface/chat.py	1	98.84%
scope3ai/tracers/huggingface/speech_to_text.py	1	98.18%
scope3ai/tracers/huggingface/translation.py	2	96.61%
scope3ai/tracers/anthropic/chat.py	4	97.08%
scope3ai/tracers/openai/chat.py	4	95.0%
scope3ai/worker.py	5	89.69%
scope3ai/api/client.py	10	81.71%
scope3ai/lib.py	17	90.23%

Totals
Change from base Build 12553763769:	15.3%
Covered Lines:	2111
Relevant Lines:	2203

---

💛 - Coveralls