avoid using local paths as "href"

[soxofaan]

in various places (mainly around batch job result/asset handling) we are putting local (absolute) paths in a "href" field. This is quite confusing as there is no indication that this is about local paths (e.g. at least a "file://" protocol would make that more explicit). It's also not ideal that at some point that "href" probably has to be overwritten with a relative/absolute (HTTP/S3) URL to be usable by end user and the original location is lost. Overall it makes it hard to reason about how assets are handled in our code

[bossie]

It's clear from the STAC spec that "href" should be a URI with a scheme so I would even call this a bug.

Re: technical debt I would also avoid passing URIs around as strings and instead use a dedicated type as it will probably get rid of a great deal of implicit assumptions and workarounds like (think // vs /).

I haven't found a dedicated URI class in Python like java.net.URI; maybe we could look into https://pypi.org/project/uri/. Automatic translation between the two by py4j should be trivial.

[soxofaan]

another symptom: #929 (comment)

[bossie]

FYI I've been playing with https://pypi.org/project/uri/ and it has some surprising behavior (marrow/uri#30 and marrow/uri#31) so doesn't look mature enough.