(some?) agents should bet on 'most recently opened' markets

[evangriffiths]

All our DeployableTraderAgents currently get markets, sorted by SortBy.CLOSING_SOONEST. This means newly created markets with a close-time more than a few days away are not traded on at all by our agents.

If some/all agents get markets sorted by SortBy.NEWEST, then they will pick up these markets and trade on them.

Considerations:

• If the number of markets that the agent (or all agents combined) fetch is < the total daily number of markets created, then some markets will still not have any trades
• Prediction accuracy will likely be lower than if making the prediction closer to market close time
• When should the agent decide to re-assess its position on the market? (see Add cheap way of deciding whether to re-run predictions for long-running markets #487)

[gabrielfior]

I think this is a duplicate of #411
We could implement different market-picking strategies on #411 and then have one (or multiple) strategies per agent, and it loops through them on each run.
@evangriffiths what do you think?

[evangriffiths]

#411 is about refactoring, this is about changing the implementation of agents. There might be one PR that solves both issues, but I don't think it's a duplicate issue.