Read.UnboundedSourceAsSDFRestrictionTracker doesn't use cache for readers in getProgress

https://github.com/apache/beam/blob/master/sdks/java/core/src/main/java/org/apache/beam/sdk/io/Read.java#L963

In getProgress, if currentReader == null a new reader is constructed from the restriction checkpoint. I think that we should consult the reader cache as creating readers is sometimes expensive.

I observed in Nexmark pipelines that readers were being created at this location.