sycamore importer: detect insufficient memory or unimportable files and abort/skip them

[eric-anderson]

The current importer will just keep retrying (with exponential backoff) when an import attempt fails. However, if the problem is insufficient memory or a document that causes the importer to crash, this is not a useful strategy.

Most likely after a few attempts to import in parallel fail, the importer should start importing documents serially and then mark ones that fail to import a few times as permanent failures and skip them until the importer code or document are updated.