[meta] Feedback from migrating to 2.0

[k88hudson-cfa]

I recently looked at migrating some existing code and had some thoughts about the migration (https://github.com/cdcent/cfa-evd-2025-interventions/pull/8)

Higher priority stuff:

• Split Property/IndexableProperty #909 impl_property! no longer works with f64 since it requires Hash, Eq, which is particularly a problem for structs that contain f64 (e.g., times for things)
• Make IxaError compatible with anyhow #908 IxaError isn't compatible with anyhow because it doesn't implement Send/Sync
• IxaError is, in practice, used quite a bit with the string variant: do we want to just add back support, or encourage people to move to defining their own errors/anyhow/something else? (we decided we will leave it but make it compatible with anyhow)
• Global property validation error should be less specific #910 Global property validation should probably take a more generic error (with Into etc.)

Other things that I thought are worth mentioning but maybe not immediately fixing

• accessing newtype property values with .0 is not ideal
• query_result_iterator + collect() does actually have use cases (particularly in reports) (we decided this was acceptable)

[RobertJacobsonCDC]

We changed the error returned by GlobalProperty::validate to be a Box<dyn Error> in #811:

The fix is to have GlobalProperty::validate return a Box. This error value is then wrapped in IxaError::IllegalGlobalPropertyValue in ixa code like GlobalPropertiesDataContainer::set_global_property_value that consumes and bubbles up these errors. This PR modifies that variant to hold the Box value and the name of the property as a String.