fix(read): fallback to raw bytes for non-UTF-8 reads

[em0t]

Summary

• preserve plain raw-read semantics for non-UTF-8 files by falling back to raw byte passthrough instead of failing early in rtk read
• keep the fallback narrow: it only applies to unfiltered reads with no line windows and no line numbers
• leave text-oriented modes unchanged, so filtered reads and -n still require valid UTF-8
• add regression coverage for the binary fallback gate

Minimal Cases

non-UTF-8 raw read

Raw

/bin/cat nonutf8.bin >/dev/null
exit 0

Previous RTK

rtk read nonutf8.bin >/dev/null
stderr contains: stream did not contain valid UTF-8
exit 1

This PR

rtk read nonutf8.bin >/dev/null
exit 0

Testing

• cargo fmt --all
• cargo test test_supports_binary_passthrough -- --nocapture
• cargo test --all
• cargo clippy --all-targets (existing repo warnings only; none introduced by this PR)