Fix problems with RA, dec sampling

[jeremykubica]

There were a few bugs found when debugging @mi-dai 's notebook that this PR fixes:

1. We should use the OpsTable's radius for RA, dec sampling unless a specific alternative is provided.

2. The radius attribute had been moved to a dictionary entry so the assignment ztf_obstable.radius was not being used in the actual simulation. This PR adds a property that allows user to access and set ObsTable radius with .radius

[github-actions]

Before [4b2b0bb]	After [e2b330e]	Ratio	Benchmark (Parameter)
110±2μs	113±0.9μs	1.03	benchmarks.TimeSuite.time_sample_x0_from_distmod
118±9μs	121±2μs	1.02	benchmarks.TimeSuite.time_evaluate_simple_linear_wavelength_model
694±20μs	711±20μs	1.02	benchmarks.TimeSuite.time_fnu_to_flam
547±20μs	554±10μs	1.01	benchmarks.TimeSuite.time_apply_passbands
19.4±0.1ms	19.5±0.2ms	1.01	benchmarks.TimeSuite.time_make_and_evaluate_static_sed
47.1±0.5μs	47.7±0.5μs	1.01	benchmarks.TimeSuite.time_make_new_salt3_model
30.7±1μs	30.9±0.7μs	1.01	benchmarks.TimeSuite.time_make_simple_linear_wavelength_model
1.02±0s	1.04±0.01s	1.01	benchmarks.TimeSuite.time_make_x1_from_hostmass
34.1±0.4ms	34.0±0.1ms	1	benchmarks.TimeSuite.time_additive_multi_model_source
19.3±0.3ms	19.3±0.2ms	1	benchmarks.TimeSuite.time_chained_evaluate_sed

Click here to view all benchmarks.