Add support for a default aliases file

[JamesWrigley]

Following up on #367 (comment), I've added support for an aliases argument to open_run() that defaults to looking for a file named usr/extra-data-aliases.yml in the proposal. So as long as that file exists, you get aliases by default. Let the bikeshedding begin! 😀

Things to mention:

• I figured that it only made sense to add an aliases argument to open_run() and not RunDirectory because this automatic-aliases-finding thing is specifically for opening runs by proposal. By that argument you could say that the aliases argument should only support paths and not dict objects, but I figured it could be useful. I have no objection to removing support for dict arguments though.
• One of .with_aliases() possible inputs is an *arg of tuples, but I didn't think that API made sense for open_run() so I left it out.

b29ad11 and b10e82c are smaller fixes that I found useful.

[philsmt]

Thanks!

Only a few nitpicks implementation-wise. While thinking about the potential silentless of this feature, I was wondering if we should just list all aliases in DataCollection.info(), rather than attach it to sources/keys if they happen to be shown (by default not the case for keys).

[JamesWrigley]

While thinking about the potential silentless of this feature, I was wondering if we should just list all aliases in DataCollection.info(), rather than attach it to sources/keys if they happen to be shown (by default not the case for keys).

What if it logged something when an alias is loaded? I'm mildly against listing all the aliases because .info() can be pretty long already.

[JamesWrigley]

Ugh, so beautiful ❤️

[philsmt]

While thinking about the potential silentless of this feature, I was wondering if we should just list all aliases in DataCollection.info(), rather than attach it to sources/keys if they happen to be shown (by default not the case for keys).

What if it logged something when an alias is loaded? I'm mildly against listing all the aliases because .info() can be pretty long already.

That is correct. I was thinking of warnings/logs too, but wasn't sure people will actually see it. Do we mind just printing something in open_run?

[JamesWrigley]

Added a print() statement in 871512b, and squashed all the fixup commits.

[philsmt]

Neat, LGTM