Retry and timeout configuration for REST catalog

[skrydal]

Feature Request / Improvement

Problem statement
REST Catalog client uses requests library to perform HTTP calls. Currently, if there is an intermittent network/server problem resulting in an error code being returned, pyiceberg library would not retry. This is a common problems for out-of-box requests usage. Also timeout for the calls is not set.

Solution outline
Since retrying/timeout functionalities are very much needed for any HTTP clients, there are classes, available off-the-shelf, which can be used:

• urllib3.util.retry.Retry - to define strategy
• requests.adapters.HTTPAdapter - to use the strategy

HTTPAdapter object can be mounted by request's session.
Timeout could be provided directly to the http calls made by the session, but even better HTTPAdapter is commonly extended, to include the timeout parameter in all calls.

There would need to be couple of new parameters provided to the catalog config. They could look as below:

       type: 'rest'
       uri: "http://localhost:8181"
       connection:
         retry:
           backoff_factor: 0.5
           total: 3
           status_forcelist: [500]
         timeout: 120

retry key value could be fed directly into the __init__ of urllibs Retry class.