Execution timeout metadata for each notebook

In light of the discussion in this thread https://github.com/jupyter/nbclient/issues/45 , we would like to know what the community thinks about storing the information of notebook level timeout for execution in the metadata so that it can be utilized by downstream tools. As sometimes we have that one notebook, which takes way too long to execute and we don't want the default timeout to be high.

There is an `execution` key for cell-level metadata, whose dict can have an additional `timeout` key added?
Or there can be a provision to store it somehow in the notebook-level metadata?

Thoughts?

cc : @choldgraf @MSeal

Start by reading the discussion in jupyter/nbclient#45 and compare the existing cell-level `execution` metadata with notebook-level metadata. Determine where a notebook-wide timeout should be represented and what downstream tools need; done requires a resolved metadata proposal rather than an open choice between alternatives.

Written by the indexing model from the issue text.