Add arabic normalizer for search

[ivanPyrohivskyi]

To issue #11709

[vshcherb]

Very ineffective too many calls for comparision plus overhead of 10 replaceAll methods

[sonora]

Yes, exactly. Some hints:

• Simplify the Kashida replacement by using 'replace' instead of 'replaceAll', more efficient (when no regex is needed).
• Use a single single map-based lookup to reduce the redundancy of multiple 'replace' calls
• Use a Stream to initialize the character replacements, makes the code easier to maintain.

[sonora]

How's this:

public class ArabicNormalizer {

    private static final Pattern DIACRITICS_PATTERN = Pattern.compile("\\p{Mn}");

    private static final Map<String, String> CHAR_REPLACEMENTS = Stream.of(new String[][] {
        {"إ", "ا"}, {"أ", "ا"}, {"ئ", "ي"}, {"ؤ", "و"},
        {"آ", "ا"}, {"ى", "ي"}, {"ة", "ه"}
    }).collect(Collectors.toMap(data -> data[0], data -> data[1]));

    public static String normalize(String text) {
        if (text == null) {
            return null;  // Handle null input
        }

        String normalized = Normalizer.normalize(text, Normalizer.Form.NFD);
        normalized = DIACRITICS_PATTERN.matcher(normalized).replaceAll(""); // Remove diacritics efficiently

        // Hamza and other normalizations
        for (Map.Entry<String, String> entry : CHAR_REPLACEMENTS.entrySet()) {
            normalized = normalized.replace(entry.getKey(), entry.getValue());
        }

        // Kashida
        normalized = normalized.trim().replace("\u0640", ""); // Kashida

        return normalized;
    }
}