Issue fixes

[jeremyrcoyle]

Omnibus PR that fixes (or attempts to fix) a number of issues:

• pipeline/stack issue: Pipeline cannot be passed to Lrnr_sl "as is", needs to be wrapped up in a list #81 Fixed by ensuring stacks/SL work with only one learner possibly themselves stacks or pipelines with stacks.
• rpart stack issue: Lrnr_rpart$train() works, but as part of stack fails #230 I added a test, but couldn’t reproduce the reported issue.
• dbarts save issue: Saving/Loading Lrnr_dbarts does not work #234 This seems to be underlying issue in dbarts, made an issue there.
• Lrnr_sl library accessor: Easier access to SL library fits #102 Added a fit$learner_fits accessor to Lrnr_sl, as well as Stack and Pipeline. Made sure the resulting learner_fits lists were named as learners are named
• Lrnr_sl computation status: optionally print computation status #181 . the sl3.verbose option now prints a message each time a learner is trained. This should help track training of compound learners