fix(temporal): prevent panic on multi-byte UTF-8 queries

[majkelooo]

Summary

• find_iso_date_in_query() in src/temporal.rs iterates over byte indices (bytes.len()) but then slices &str using those indices (&query[i..i + 10])
• When the query contains multi-byte UTF-8 characters (e.g. Polish diacritics: ą, ś, ź, ż — each 2 bytes), a byte index can land inside a character, causing a panic at runtime
• Added is_char_boundary() guards before slicing to skip non-boundary positions

Reproducer

engraph search "powiązania między klientami"
# panicked at src/temporal.rs:279:31:
# byte index 5 is not a char boundary; it is inside 'ą'

Any query containing Polish, Czech, German (ü, ö), or other non-ASCII characters at specific byte offsets triggers this.

Fix

for i in 0..=bytes.len() - 10 {
    if !query.is_char_boundary(i) || !query.is_char_boundary(i + 10) {
        continue;
    }
    let candidate = &query[i..i + 10];

Test plan

• engraph search "powiązania między klientami" — was panic, now returns results
• engraph search "księgowość rozliczenia" — was panic, now returns results
• engraph search "świadczenie usług" — was panic, now returns results
• ASCII queries still work as before
• cargo build --release passes

🤖 Generated with Claude Code