Tolerate stream metadata that is not valid UTF-8

[adityasingh2400]

StreamInfo.hostname() raises UnicodeDecodeError on Windows machines whose hostname contains accented characters, as reported in #69. as_xml() fails in the same way because the hostname is embedded in the info document, so a resolver loop can die simply from discovering a stream that lives on such a machine.

The root cause is that every string coming back from liblsl was run through a strict bytes.decode("utf-8"). That is safe for values pylsl itself wrote, since pylsl encodes them as UTF-8 on the way in, but the hostname is read from the operating system and Windows hands it back in the active ANSI code page. The same applies to any metadata written by a non-Python LSL application running on such a machine.

This adds a small _to_str helper in pylsl.util that tries UTF-8 first and falls back to latin-1. The fallback cannot fail, keeps every original byte so a caller can re-encode and decode with the correct code page if it knows better, and renders the common accented-Latin case correctly, which is the fix the reporter validated in the issue thread. The helper is applied to the metadata accessors in info.py: name, type, source_id, uid, session_id, hostname, as_xml, and the XMLElement name, value, and child_value getters. The sample data path in inlet.py is deliberately left alone, since raw bytes in cf_string channels are a separate discussion in #54.

Tested with a new test/test_encoding.py that exercises the helper directly and simulates a non-UTF-8 platform by monkeypatching the relevant liblsl calls to return cp1252 bytes. With the tests in place but info.py reverted, three of them fail with exactly the error from the issue, UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe1 in position 7. With the change applied the whole suite passes, 18 tests against liblsl 1.17.7 on macOS with Python 3.14. ruff check and ruff format --check report nothing new relative to main.

Fixes #69