Improve Dependency Reproducibility Robustness

In four sections going from "we are asking for disaster if don't do this" to "wow, that would be pretty cool".

For the community on depot for we should back up all:

- [ ] Conda artifacts used by known tools. https://github.com/maxpoint/conda-mirror
- [ ] Quay.io containers under the biocontainers namespace.
- [ ] Singularity containers matching the biocontainers.

On Galaxy instances we should record:

- [ ] Record the full list of packages (including versions and sources) used in Conda for a given job.
- [ ] Record every container identifier and some sort of hash.
- [ ] Provide instructions and potentially infrastructure for local Docker repositories to store and serve containers.

On Galaxy instances we should provide the option to backup:

- [ ] All Conda packages used for every job.
- [ ] All containers used for every job.

On Galaxy instances we should provide APIs for:

- [ ] Fetching commands and package lists for rebuilding Conda environments for given jobs outside of Galaxy.
- [ ] Downloading the container used for a given job.
- [ ] Build containers with packages matching those used natively for non-container jobs.

No files, tests, or entry points are identified. Start by breaking the four sections into separately scoped work and reading the linked conda-mirror project, then define which checklist item is being implemented and how backup, recording, or API behavior would be verified.

Written by the indexing model from the issue text.