fix(sparql-anything): accept an empty output from a readable, non-empty chunk - #836

convert() rejected any process whose --output was empty, because SPARQL Anything exits 0 when it cannot read or parse an input. A chunk whose rows the query filters out entirely writes a 0-byte output too, and was then reported as an unreadable input, aborting the run.

The converter now tells the two apart by what it can see before starting a process:

• A relative chunk or load path is relative to workDir for the task runner and for the converter alike, so it is checked first: a missing or empty one fails before any JVM runs, naming the file.
• An absolute path is the runner’s – under a container’s mount, say – and one that does not exist here proves nothing, so the process runs and its output is held to the old rule.
• Once every input of a process was seen to be non-empty, an empty output is accepted as “no triples” and contributes nothing to the concatenation. A missing output still fails, as does an empty output of a job without chunks: its input is named in the query, which the converter cannot see, and the admin-codes style query has no legitimate empty result.

Docs updated under “Where files are written” and “How a conversion runs”.

Fix #820