Willingness to Pay (WTP) has many ways to compute it offered in the literature...some better than others. Sawtooth Software has an approach that is part of its market simulator: https://sawtoothsoftware.com/resources/technical-papers/estimating-willingness-to-pay-in-conjoint-analysis
You cannot tell how much people are willing to pay for a single level alone (e.g., level 1 of A2) with standard conjoint analysis (CBC). You can only tell how much people are willing to pay for one level of an attribute compared to another level of the same attribute.
You can tell how much people are willing to pay for a shift from level 1 to level 2 or 3 of that SAME attribute.
It is possible to linear interpolate between attribute levels: both for the price attribute and the other attributes.
With dummy-coding, you can linear interpolate between the utilities of adjacent levels.
An attribute that has mixed quantitative and qualitative levels could by dummy-effects coded. Or, I think a fancier coding approach could offer a mixed approach (that dummy-coded the first level and linear coded the quantitative levels. That would be just a twist on the "piecewise coding" approach.