Percent-encoding: add optional iri feature for IRI-style UTF-8 in encode output

[fernandolins]

Context

This change is motivated by the discussion in #871 (readability of URLs when non-ASCII characters are percent-encoded, and making that behavior configurable without breaking the default).

Summary

Adds a Cargo feature iri on the percent-encoding crate. When enabled, AsciiSet::should_percent_encode only considers ASCII bytes: non-ASCII UTF-8 octets are not forced through percent-encoding based solely on being non-ASCII. When disabled (default), behavior matches the existing WHATWG-aligned rule (!byte.is_ascii() || self.contains(byte)).

Motivation

Some consumers (e.g. local file: / media paths in HTML) want Unicode segments to remain literal in the encoded string rather than always escaping every UTF-8 byte. That differs from the current default, which encodes all non-ASCII octets. Upstream url and WPT expectations rely on the default, so this is opt-in via a feature flag rather than a breaking change.

Behavior

iri	Non-ASCII UTF-8 bytes	ASCII bytes in the set
off (default)	Percent-encoded	Encoded
on	Passed through	Encoded

API / compatibility

• Default: unchanged for existing dependents and for url + conformance tests.
• Opt-in: percent-encoding = { version = "…", features = ["iri"] }.

Testing

• cargo test (workspace default, no iri on percent-encoding in url’s manifest).
• cargo test -p percent-encoding --features iri.