Reconcile unicode characters in record attributes/searchable terms

We handle non-ASCII characters in database records by converting them to HTML escape sequences:

(however github's markdown renderer will display them as unicode -- the actual text of the value below is `&` folloewd by `alpha` followed by `;`)

> "label": "CXCL12α",

http://normalize.cancervariants.org/therapy/normalize?q=pubchem.substance:135651881

However, these values aren't properly handled either in the SwaggerUI interface (the label value above is converted to `CXCL12&alpha`) or via an HTTP request (the label value above gets the escape sequence cut entirely and the final query is `CXCL12` which is an entirely different drug)

Reproduce the reported CXCL12α example through the linked normalization request and the SwaggerUI interface, comparing the returned label and final query. Trace where record attributes and searchable terms are converted or parsed, then verify that the complete value survives both paths without changing the intended search term.

Written by the indexing model from the issue text.