[old doclet] Javadoc search index generation is not thread safe

AbstractIndexWriter.createSearchIndexFile creates unzipped json file, zips it and then deletes. If another thread deletes same unzipped index in the middle, the original one fails with NoSuchFileException on delete part.

Contents of the file may be just in memory or there may be a file-locking or just synchronization to make entire operation atomic.

Actual practical scenario is multi-threaded benchmarking so not sure if it is really an issue but there were no exceptions in the past.