Scml

-----------------

* [bugfix] Bankrupt agents kepts requesting negotiations
* [bugfix] trading prices inaccurate during step
* [std] Adding spot_market_quantity/loss to the AWI
* [oneshot] Adding helpers to avoid crazy prices

-----------------

* [std/collusion] keeping compatibility with 2020
* [std/collusion] Avoiding crash when agent class names is shorter than 2
* letters
* [std/collusion] rare division by zero error
* [oneshot] forcing n_processes to 2 in tournaments

-----------------

* [all tracks] Adding convenience methods to AWIs. closes 49
* [all tracks] reducing production cost range
* [all tracks] casting offers to ints explicitly to avoid bugs in agents offering fractional quantities/unit_prices
* [oneshot] avoiding ultimatum in oneshot (was disabled)
* [oneshot] Add `current-inventory` to OneshotAWI which will always return zero for compatibility with the other tracks
* [onesht] adding public_* to summary.
* [oneshot] penalty scale was incaccurate
* [std/collusion] Adding satisficer agent
* [std/collusion] adding guarnteed_profit method of world generation
* [std/collusion] better handling of predictions in builtin
* [std/collusion] experimental better decentralizing agent
* [std/collusion] Increasing profit potential
* [speedup] avoid saving negotiations online

-----------------

* [oneshot] refactoring using before_step in greedy
* [minor] formating update
* [bugfix] production costs were not increasing. closes 38
* [API] Adding before_step() to all agents. If a method called `before_step()` is defined for an agent, it will be called once every simulated day before any other calls to the agent but after all exogenous contracts and ufun parameters are fixed. Note that `step()` is called at the *end* not the beginning of the day.
* [oneshot] limiting exogenous quantities to n lines. This closes 37
* [oneshot] Aspiration negotiator gets more agreements
* [oneshot] allowing agents to skip their turn. Agents can skip their turn now by returning REJECT_OFFER, None
* [oneshot] GreedyOneShotAgent is more rational now
* [bugfix] avoiding an inconsistency in path names for logs
* [oneshot] improved builtin agents (aspiration)
* [tournament] avoiding a possible edge case that would have led to competitor agents appearing as non-competitors
* [tournament] Saving negotiatinos by default in the CLI and utils
* [std/coll] defaulting to narrower worlds

----------------

* [oneshot] Matching default parameters of world generation to game description.

----------------

* [oneshot] Adding OneshotIndNegotiatorsAgent to use independent negotiators in oneshot
* [std/coll] updating builtin compoenents to be more rational
* [bugfix] failure in distributing products when the number of agents per process becomes large.
* [cli] changes on default competitors in the CLI
* [oneshot] adding the option to disable avoid-ultimatum (and disabling it for now)
* [oneshot] making sure needs are integers in greedy
* [bugfix] negotiator id is not the same as partner id sometimes
* [cli] adding --name to run2021 command to control world name
* [bugfix] negotiator_id and partner_id were not equal
* [eval] ensuring that ageents are run in exactly the same conditions
* [eval] adding zscore, iqr, fraction (old iqr -> iqr_fraction) to truncated mean
* [cli] changing default n. competitors to 2
* [oneshot] adding an option not to count/not count future contracts on bankrruptcy