Add configuration parameter to limit stream size read from disk.

[jmchilton]

Set with maximum_stream_size in app.yml.

Default to 1MB.

Fixes underlying problem in #151 I think, though doesn't actually do what was requested. I prefer this - is that okay @natefoo?

[coveralls]

Coverage increased (+0.1%) to 82.431% when pulling b4b4aa2 on jmchilton:output_limit into afd5886 on galaxyproject:master.

[coveralls]

Coverage increased (+0.1%) to 82.431% when pulling b4b4aa2 on jmchilton:output_limit into afd5886 on galaxyproject:master.

[coveralls]

Coverage increased (+0.1%) to 82.431% when pulling b4b4aa2 on jmchilton:output_limit into afd5886 on galaxyproject:master.

[natefoo]

It'd be nice if we could avoid losing the stdout/stderr as the error message you'd use to determine why a job failed is likely at the end. But this will at least avoid the biggest problem, which is RabbitMQ crashing on huge messages.

[jmchilton]

It'd be nice if we could avoid losing the stdout/stderr as the error message you'd use to determine why a job failed is likely at the end.

Agreed.

And another next step that would be nice from an API perspective is to avoid the concatenation that Galaxy does. I'd rather have structured information - like {..., "stdout_contents": "...", "stdout_truncated": true, "stdout_contents_end": "..."} in this API and then let Galaxy mangle the data I think.

[natefoo]

What's the concatenation?