Allow indifference in ordinal elicitation

[gertvv]

In ordinal elicitation, allow multiple answers to the elicitation question (which means indifference between those choices). The preferences translate as follows:

• Say c1 > (c2 ~ c3) > c4, then add c1 > c2, c1 > c3, c2 > c4, c3 > c4 as constraints.

Ratio bound elicitation becomes slightly more complex. These cases must be taken into account:

• (c2 ~ c3) specifies an indifference class. Within this class, the weight rations w2/w3 must include 1. This means that two separate questions must be asked to obtain the upper and lower bounds for the ratio, otherwise the slider would need to go outside the range where the PVFs are defined.
• If c1 > (c2 ~ c3), then we could ask ratio questions about w1/w2 or w1/w3, neither contain exactly the same information; should this choice be made arbitrarily, or should all questions be asked?
• Within a larger indifference class (e.g. c1 ~ c2 ~ c3), n(n-1)/2 questions can be asked, and again there is no "natural" ordering to the questions.

Some of these could be solved by the entropy-minimization approach, but that is not finished.