Default`n_adapts` and `discard_initial` to zero

[yebai]

Some downstream MCMC sampling package, e.g., AHMC, now assumes we always pass n_adapts to the AbstractMCMC.step function. It might be sensible to default these arguments to 0 when users didn't specify them to avoid missing n_adapts errors.

[codecov[bot]]

Codecov Report

Patch and project coverage have no change.

Comparison is base (d7c549f) 97.37% compared to head (8dc2dcf) 97.37%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #124 +/- ## ======================================= Coverage 97.37% 97.37% ======================================= Files 8 8 Lines 305 305 ======================================= Hits 297 297 Misses 8 8 ``` | [Files Changed](https://app.codecov.io/gh/TuringLang/AbstractMCMC.jl/pull/124?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=TuringLang) | Coverage Δ | | |---|---|---| | [src/sample.jl](https://app.codecov.io/gh/TuringLang/AbstractMCMC.jl/pull/124?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=TuringLang#diff-c3JjL3NhbXBsZS5qbA==) | `96.64% <ø> (ø)` | |

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.