Avoid repeated normalization of every alias pair in `diffPlaces`

While testing address documents with many name aliases, duplicate checking became much more expensive than the Elasticsearch lookup. proposing a small optimization that preserves the existing duplicate decisions.

isPropertyDifferent compares every value on one side with every value on the other, normalizing both inside the nested loop. For disjoint arrays of lengths m and n, that performs O(m × n) normalization work for each property comparison.

In a NYC address result set I was using in local testing, 19 of 20 documents had 169 additional name aliases each. Comparing all 190 document pairs through isDifferent produced:

The proposed change is to normalize one array into a set, then normalize each value in the other array once and test membership. then return “same” on any overlap, retaining the current missing property and scalar to array handling. This makes the property-level comparison O(m + n), with O(n) temporary storage.

Start in helper/diffPlaces.js at isPropertyDifferent, especially lines 256-276, and trace its callers through isDifferent. Preserve the existing missing-property, scalar-to-array, and duplicate-decision behavior while avoiding repeated normalization; compare the result against the reported 190-pair behavior and performance improvement.