The backtrace and context of the original error is lost
when we rescue and raise a new error. This makes it difficult
to debug.

If a report has been tampered with and is invalid
it is safer to return an empty report than to
bubble up an error that cannot be dealt with
meaningfully by clients of the parser code.

There is no action that we can take
when a JSON parse error occurs due to
an invalid JSON document. Re-processing
the document will not yield a different result.

It is very easy to upload a `gl-license-scanning-report.json`
that is not a valid JSON document. This change stops recording
errors that we cannot take action on.