Content app shouldnt retry on RA streaming

[pedro-psb]

Our RemoteArtifact streaming uses chunked transfer, so once it has started a response, we can't fix anything anymore in case an error.

Because of that, the retry logic (any of the timeout, digest, etc) on the http downloader is just delaying a response that can't be possible right anymore.

We should add a mechanism to skip retries when calling from the chunked streaming context.

Relevant code:

• Retries on downloader:

  pulpcore/pulpcore/download/http.py

  Lines 230 to 250 in 24dac1b

  	retryable_errors = (
  	aiohttp.ClientConnectorSSLError,
  	aiohttp.ClientConnectorError,
  	aiohttp.ClientOSError,
  	aiohttp.ClientPayloadError,
  	aiohttp.ClientResponseError,
  	aiohttp.ServerDisconnectedError,
  	TimeoutError,
  	TimeoutException,
  	DigestValidationError,
  	SizeValidationError,
  	)
  	async with self.semaphore:
  	@backoff.on_exception(
  	backoff.expo,
  	retryable_errors,
  	max_tries=self.max_retries + 1,
  	giveup=http_giveup_handler,
  	)

• Download call from stream_remote_artifact:

  pulpcore/pulpcore/content/handler.py

  Lines 1119 to 1126 in 24dac1b

  	downloader = remote.get_downloader(
  	remote_artifact=remote_artifact, headers_ready_callback=handle_response_headers
  	)
  	original_handle_data = downloader.handle_data
  	downloader.handle_data = handle_data
  	original_finalize = downloader.finalize
  	downloader.finalize = finalize
  	download_result = await downloader.run()

Context

Identified this when investigating #5012 and #5725