Make scraper dependency optional or replace with lightweight alternative

[digitalcortex]

I'm using impit as a library for browser fingerprinted HTTP requests. Impit making scraper crate a required dependency causes two problems:

First problem: namespace collision. Scraper is a common crate name. When impit pulls it in as a transitive dependency, any workspace that also has a crate named scraper gets ambiguous resolution, for example I had to specify exact version of my own crate in order for this command to run: cargo run -p scraper@0.1.0 --bin scraper --, otherwise I was getting an error:

error: There are multiple `scraper` packages in your project, and the specification `scraper` is ambiguous.

Since scraper is a generic name, this is likely to affect other consumers too.

Second problem: using scraper as dependency is too heavyweight for what it does. The only usage of scraper crate I found is in response_parsing/mod.rs — prescan_bytestream() parses the first 1024 bytes into a full html5ever DOM tree just to read and attributes. This pulls in html5ever, selectors, ego-tree, cssparser, etc. for a two-selector lookup.

Additionally, this encoding detection isn't even wired into the response pipeline — Impit::send() returns a raw reqwest::Response. The decode() function is only exposed as an opt-in utility via impit::utils, so consumers who don't call it still pay the compile cost.

Suggested fixes (either would work):

• Make scraper optional behind a feature flag (e.g. encoding-detection) so consumers who don't use impit::utils::decode don't pull it in.
• Replace "scraper" with lol_html, a streaming HTML parser that can extract charset from tags without building any node trees. Different name avoids the namespace issue, and it's far lighter for this use case.