When anchoring many annotations on a PDF, the text of the same page may
be requested many times.

The text is initially fetched via an async call to
PDFViewerApplication.pdfViewer.getPageTextContent() and the result is
then cached for future use.

Previously the cache was only updated once this expensive call
returned a result. Consequently when anchoring many annotations at once,
there would be a lot of unnecessary cache misses and repeated calls to
PDFViewerApplication.pdfViewer.getPageTextContent()

Fix the problem by storing a promise for the result in the cache when
the text for a given page is requested the first time. Consequently the
text for each page will only be fetched once.