Fix some parsing issues with complex truth catalog entries

[rmjarvis]

@NiallMac was having some problems getting some values into the truth catalog for DES runs. We eventually found a workaround, but this PR fixes the underlying issues to make it easier to get these kinds of items parsed properly.
cf. https://github.com/des-science/y3-wl_image_sims/issues/82

Specifically, I've made the following changes:

1. When an '@' item appears in an Eval string, and the item hasn't been evaluated yet (so the config processing, doesn't know what type it is), then if there is only one possible type, go ahead and assume that's the type to use.
2. For types that allow multiple value types (e.g. List, Random, etc.), automatically add parallel types that look like e.g. Random_float, Random_Angle, etc. which can utilize the above solution.
3. Don't consider np.float64 and float to be different types when building the truth catalog. All float-y types are now float, and all int-y types are int.
4. When replacing the @ values in an Eval string, do so in reverse sorted order to make sure e.g. @gal.index is replaced before @gal.
5. Added better error messaging about mismatched types to make it easier to know where the problem is. (It had been fairly opaque.)
6. Added a suggestion to use e.g. Random_float when it can't figure out what type a Random should be when it evaluates in a context where it doesn't know yet.

[rmjarvis]

@NiallMac Do you mind taking a look at this? I think it solves all of the problems you'd been having with the truth field.

You don't have to check the implementation, but if you can at least look at the unit tests and make sure I hit all the things we tried that we thought should have worked but didn't, that would be much appreciated.

[rmjarvis]

Any objection to merging this? I have more config-layer work I've been working on, so I'd like these changes to get onto master before that PR. I'll merge Friday unless I hear that anyone wants more time to review this.

[NiallMac]

Sorry for the delay on this Mike, just got to it now. The expanded unit test looks great to me.