feat: support two-step OpenSearch Description for OPDS search

Servers like copyparty advertise search via a link to an OpenSearch
Description (OSD) document rather than embedding {searchTerms} directly
in the feed's <link rel="search"> href (the Calibre-Web style).

OpdsParser now stores the OSD URL when rel="search" and
type="application/opensearchdescription+xml" but the href lacks
{searchTerms}. OpdsBookBrowserActivity fetches that OSD document after
each feed load and parses the <Url template="..."> element with a
minimal inline expat parser to extract the actual search template.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

lib/OpdsParser/OpdsParser.cpp

@@ -70,6 +70,7 @@ bool OpdsParser::error() const { return errorOccured; }
 void OpdsParser::clear() {
   entries.clear();
   searchTemplate.clear();
+  osdUrl.clear();
   nextPageUrl.clear();
   prevPageUrl.clear();
   currentEntry = OpdsEntry{};
@@ -105,6 +106,8 @@ void XMLCALL OpdsParser::startElement(void* userData, const XML_Char* name, cons
         std::string sHref(href);
         if (sHref.find("{searchTerms}") != std::string::npos) {
           self->searchTemplate = sHref;
+        } else if (type && strcmp(type, "application/opensearchdescription+xml") == 0) {
+          self->osdUrl = sHref;
         }
       } else if (rel && strcmp(rel, "next") == 0 && !self->inEntry) {
         self->nextPageUrl = href;