perf(java): reuse usage search results and skip covered folders when widening (#2278)

Find Usages and usage counts searched again on every request: about two
seconds each for a common type such as String.

- The Java usages extension (1.12.0) keeps answers by search target until a
  Java element, the classpath or an unsaved edit changes, and requests for
  the same target share one search.
- Widening checks up to 5000 candidates against the server's analysis scope
  before applying the 300-file limit, so names mentioned in many covered
  files no longer report "too many files".
- Candidate scans skip the source folders the server covers fully
  (openidea.workspace.coveredRoots) instead of reading every file.
- Commands a server registers after initializing are added to its commands.

java/java-usages/README.md

@@ -27,6 +27,11 @@ from directory names. References, rename and incoming calls leave those files cl
 and only widen with the remaining candidates. A missing or failing command retains
 widening. The result is query-local, so imports and classpath changes need no cache invalidation.
 
+`openidea.workspace.coveredRoots()` returns the URIs of source folders whose Java files the search
+scope covers all: no exclusion patterns, no inclusion pattern other than Maven's `**/*.java`, and
+enclosed by the test-source policy. The client does not read the files under them when it looks
+for candidates, which on a large project took most of a cached search's time.
+
 ## Search plan and access semantics
 
 For source fields on the verified JDT LS implementation, a single plan merges
@@ -58,7 +63,14 @@ traversals. It indexes only requested exact AST ranges and verified invocation
 selector-to-call-end spans; duplicate, recovered, malformed or uncontained ranges
 fall back to NodeFinder. Method references do not get invocation aliases. Source
 snapshot checks remain per result, and the index is discarded after the query.
-There is no cross-request cache or binding-resolution shortcut.
+
+Answers are cached by their search targets (JDT handles, kinds, declarations), the scope and the
+search settings, together with a digest of the editors' unsaved contents, until a Java element
+delta reports a change to the sources, a compilation unit or a classpath (`UsageCache`). Opening
+or closing an editor and decompiled library classes' private working copies do not invalidate
+them. A usage count and the usages shown when it is clicked, or the same search run again, need
+no second search, and concurrent requests for the same targets share one. There is no
+binding-resolution shortcut.
 
 Popup and Find share persisted read/write toolbar switches. Both initially show;
 turning the last one off automatically enables the other. Unknown and readWrite
@@ -76,7 +88,7 @@ is not a claim of full IntelliJ feature parity.
 ## Rebuild and delivery
 
 The checked-in jar is embedded by ide-sdk `include_bytes!`, installed as
-`java-usages-1.9.0.jar`, and added alongside existing initialization bundles.
+`java-usages-1.12.0.jar`, and added alongside existing initialization bundles.
 Users need no compiler, Maven or Python. Manifest/install version must change
 when replacing the bundle because OSGi caches bundles.
 
@@ -118,6 +130,9 @@ longer resolves every Spring Boot property again; its reproducible SHA-256 is
 Version 1.9.0 adds the analysis-scope command to avoid redundant semantic queries;
 its reproducible SHA-256 is
 `ae04182d6a86dfdd6f6befaaeb28907fd9e4c022ceedb445f0b32ea917dca379`.
+Version 1.12.0 caches usages answers until the Java model or an unsaved edit changes and adds
+`openidea.workspace.coveredRoots`; its reproducible SHA-256 is
+`c1bd28d7b956cb64af3e936aac0553413a5dc5fd373a043ddba003009e0ea783`.
 
 ```sh
 export PATH=/path/to/jdk25/bin:$PATH