Cap merged Document search results at the requested limit

[raphyabak]

Summary

Fixes #532.

Document.search() with { merge: true } runs each field's search independently with the given limit, then merges the per-field results by unique id. When different fields match different, non-overlapping sets of ids, the merged total can exceed limit by up to (number of fields × limit), since nothing re-applies the cap after merging:

const index = new FlexSearch.Document({
    document: {
        id: "id",
        index: ["name_1", "name_2", "name_3"]
    }
});

for (let i = 0; i < 30; i++) {
    index.add({ id: i, name_1: names[i % 2], name_2: names[(i + 1) % 2], name_3: names[i % 2] });
}

index.search("foo", { merge: true, limit: 10 }).length; // 20, not 10

Fix

The overall limit requested by the caller is captured before the per-field loop can override it for individual fields, and merge_fields() now stops collecting once that many unique results have been found.

Test plan

• Added a test in test/document.js reproducing the reported case (three fields matching non-overlapping id sets, merge: true, limit: 10), asserting the merged result length is exactly 10.
• Verified the regression: reverting the source change turns the new test from passing (10) into a failure (20), confirming it actually catches the bug.
• Full suite passes locally: npx mocha test/*.js --exit — 129 passing, 1 pending (pre-existing, unrelated to this change).