Lmo

Bugfix release: fix `cache=True` if `r` is larger than what's cached

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.6.0...v0.6.1

- Optional `cache=True` kwarg to speed up consecutive L-moment calculations

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.5.3...v0.6.0

Fixed broadcasting in `lmo.l_ratio`

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.5.2...v0.5.3

- `lmo.diagnostic.normaltest`: robust test for normality (7)
- `lmo.stats.l_moment_from_ppf`: (experimental) calculation of population (theoretical) L-moments from a quantile function (5 )

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.5.1...v0.5.2

- Non-parametric exact (T)L-moment covariance matrix: `[lmo.l_moment_cov`](https://jorenham.github.io/lmo/reference/#lmo.l_moment_cov) 4
- Non-paramteric exact-ish (T)L-moment ratio standard error: [`lmo.l_ratio_se`](https://jorenham.github.io/lmo/reference/#lmo.l_ratio_se) 4
- Internal module restructuring; more methods are public now
- Improved [docs](https://jorenham.github.io/lmo/reference/)

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.5.0...v0.5.1

* All calculations are now fully vectorized, and around ~100x faster
* Frequency weights for the univariate L-moments.
* Most methods now have usage examples.
* More and better tests.
* A bunch of minor bugfixes.
* BREAKING: Merged the `tl_{}` into the `l_{}` methods, for the sake of "DRY".
* BREAKING: The `weights` parameter is now called `aweights`.

**Full Changelog**: https://github.com/jorenham/lmo/compare/v0.4...v0.5.0