add tolerance type

[rlanzafame]

Issue #3 describes the "all or each" approach, which should also be applied to this.

allow for different types of tolerance, specified in a new dict key: tolerance_type

I propose these types, defined here, but check the numpy isclose Notes and make sure you are consistent with that:

• absolute as expected
• relative as expected

Example:

exercise = {
    "type": "values",
    "tolerance": 0.01,
    "tolerance_type": "absolute",
    "variables": ['multiplication', 'addition'],
    "values": [13.889, 7.53]
}