reduce time on zexperimental tests

[chaozg]

fixed #413

Issue: the tests on the new samplers take too much time and make the local pytest and github build very slow. Tests on RegularizedLinearRTO and NUTS are the most expensive ones. Solution: as suggested by Nicolai, we

• reduce the default length of the chain from 100 to 20
• keep only 2 of the default targets (one being dim 2 and the other dim 128)
• keep only 4 of the regularized targets for RegularizedLinearRTO ([single likelihood, multiple likelihood] x [dim 16, dim 128])

The below shows the top 15 time-consuming tests in tests/zexperimental/test_mcmc.py before and after the change:

(before) slowest durations ==============================================================================
4.82s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target5]
2.94s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target2]
2.67s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_warmup[target1]
2.52s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_sample[target5]
2.48s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_sample[target1]
2.38s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target4]
2.08s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target3]
1.86s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_warmup[target2]
1.80s call     tests/zexperimental/test_mcmc.py::test_UGLA_regression_warmup[128]
1.67s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_sample[target2]
1.59s call     tests/zexperimental/test_mcmc.py::test_UGLA_regression_warmup[64]
1.54s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target1]
1.47s call     tests/zexperimental/test_mcmc.py::test_CWMH_regression_warmup[target2]
1.29s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target0]
1.29s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_sample[target4]

(after) slowest durations ==============================================================================
1.31s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_warmup[target1]
0.95s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target3]
0.78s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_sample_tune_first_step_only[target1]
0.56s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target1]
0.55s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_sample[target3]
0.42s call     tests/zexperimental/test_mcmc.py::test_CWMH_regression_warmup[target1]
0.36s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target2]
0.33s call     tests/zexperimental/test_mcmc.py::test_MH_regression_sample[target0]
0.32s call     tests/zexperimental/test_mcmc.py::test_UGLA_regression_warmup[128]
0.31s call     tests/zexperimental/test_mcmc.py::test_checkpointing[sampler3]
0.28s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_sample[target1]
0.27s call     tests/zexperimental/test_mcmc.py::test_checkpointing[sampler2]
0.27s call     tests/zexperimental/test_mcmc.py::test_state_is_fully_updated_after_warmup_step[sampler6]
0.24s call     tests/zexperimental/test_mcmc.py::test_RegularizedLinearRTO_regression_warmup[target0]
0.24s call     tests/zexperimental/test_mcmc.py::test_NUTS_regression_sample_tune_first_step_only[target0]