Learner1D: return inf loss when the bounds aren't done

[basnijholt]

Description

@bernardvanheck found an issue that occurs when using multiple cores on the following function

Here the middle region is exponentially flat and the blue curve is a homogeneous plot and the orange points are of the finished learner with a very small loss (<0.01).

I think that a learner should report an infinite loss whenever the bounds aren't included.

Having that, means that the issue I described above is much less likely to occur.

The following code snippet results in a learner with a small loss after only 3 points.

import adaptive
from adaptive.learner.learner1D import curvature_loss_function, Learner1D
from adaptive.runner import replay_log

learner_fail = Learner1D(
    lambda x: x, bounds=(-400e-6, 400e-6), loss_per_interval=curvature_loss_function(),
)

# The first returning points are in the middle flat region.
runner_log = [
    ("ask", 4),
    ("tell", 0.00013333333333333334, 1.9998184085251902),
    ("ask", 1),
    ("tell", -0.00013333333333333334, 1.9998184085251902),
    ("ask", 1),
    ("tell", 0.0, 1.9998184085251902),
]
replay_log(learner_fail, runner_log)

runner = adaptive.BlockingRunner(learner_fail, goal=lambda learner: learner.loss() < 0.01, log=True)

Checklist

• Fixed style issues using pre-commit run --all (first install using pip install pre-commit)
• pytest passed

Type of change

Check relevant option(s).

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update