Better HdrHeap Handling

Most of the interesting data (request/response headers etc.) is stored in alternates in a _marshaled_ `HdrHeap` struct. I don't understand very well how this structure works, and the current method of extracting a URL from one them involves some assumptions and guesswork. To go further, a deeper understanding is needed (I'm looking at you, amc).

Start by tracing the current URL extraction logic and reviewing how Apache Traffic Server's marshaled HdrHeap stores request and response headers in alternates. The issue identifies no files or tests, so first map the existing assumptions and determine how amc relates to the structure. Done means URL extraction handles the relevant alternates without guesswork.

Written by the indexing model from the issue text.