Fix issue with header prefixes, remove ambiguous header processing

[zootalures]

The python FDK has ambiguous and incorrect behaviour around header processing:

• It incorrectly processes HTTP gateway headers, using lstrip instead of substring: this causes strange behaviour when an HTTP header starts with any strings matching [fnhtp-]+ and where that string is stripped instead of the actual http prefix "fn-http-h" (e.g. fn-http-h-good -> "good" but fn-http-h-hello -> "ello")
• It incorrectly collapses HTTP headers into the function headers - this makes it impossible for a user to capture the exact HTTP headers sent to an API gateway as they are intermingled with the underlying function call headers
• It ambiguously processes headers when this collapsing happens - (based on dictionary order) - so if I have two headers from functions "fn-http-h-myheader" "myheader" it may present either depending on the order the keys are evaluated in the dictionary

This fix resolves all three problems.

For the second problem, to avoid breaking an backwards compatibility I've added a ctx.HTTPHeaders() method which only returns the incoming request headers.

The testing behaviour was also ambiguous as it relies on the internal code in the FDK that is under test so I added a raw testing call that does not apply the internal transformations to headers to the test fixture.

Rather than using the function boundary to assert data I added a global capture to the fixture to retrieve the exact context presented to the function for testing.

[garthy]

Suggested change

	async def setup_fn_call_raw(handle_func, content=None, headers={}):
	async def setup_fn_call_raw(handle_func, content=None, headers=None):
	if headers is None:
	headers = {}

[zootalures]

fixed.

[reclaro]

LGTM I am wonder if we can add a small test of the http override where we highlight the precedence of http headers over functions ones

[reclaro]

nit: i guess this is a leftover from the a debug session

[zootalures]

ooh yeah
thanks

[reclaro]

Thanks for adding tests and the logic to merge headers to we don't lose any value. The logic is a bit complicated and not perfect but this PR is fixing 2 bugs so i think we should merge it. There are a couple of unecessary print left in the code that should be removed.