Add skew-lognormal (skewlnorm) distribution family

[joethorley]

Closes #127.

Adds a skew-lognormal distribution family, where $X = \exp(Y)$ and $\log(X) = Y$ follows a skew-normal distribution with location meanlog, scale sdlog, and shape.
It mirrors the existing skewnorm family and reduces exactly to the log-normal when shape = 0, just as skewnorm reduces to norm.

Added

• Base distribution functions in R/skewlnorm.R: dskewlnorm(), pskewlnorm(), qskewlnorm(), rskewlnorm() (parameterised meanlog / sdlog / shape, delegating to the sn package).
• Helper family integration: ran_skewlnorm(), sens_skewlnorm(), quant_skewlnorm(), dev_skewlnorm(), res_skewlnorm(), log_lik_skewlnorm(), prob_skewlnorm().
• Tests in tests/testthat/test-skewlnorm.R plus family blocks in the existing test files.
• _pkgdown.yml reference index and NEWS.md entries.

Implementation notes for review

• Density $f(x) = f_{SN}(\log x)/x$ for $x > 0$, $0$ otherwise; CDF $F(x) = F_{SN}(\log x)$; quantile $\exp(Q_{SN}(p))$. Each base function mirrors the edge-case handling (NULL / character / NA / length / sdlog checks) of R/skewnorm.R.
• dev_skewlnorm() mirrors dev_lnorm() (log-transform then dev_skewnorm()).
• res_skewlnorm() and sens_skewlnorm() use the closed-form skew-lognormal moments derived from the skew-normal MGF: $E[X]=2e^{\mu+\sigma^2/2}\Phi(\delta\sigma)$, $E[X^2]=2e^{2\mu+2\sigma^2}\Phi(2\delta\sigma)$, with $\delta=\alpha/\sqrt{1+\alpha^2}$ (verified against simulation).
• Design decision to confirm: sens_skewlnorm() preserves the natural-scale mean and scales the SD by sd_mult via a 1-D root solve (holding shape fixed), consistent with sens_lnorm() and reducing to it at shape = 0. Flagging in case a log-scale adjustment is preferred instead.

All affected test files pass locally (sn-gated tests skipped when sn is unavailable).

🤖 Generated with Claude Code

[aylapear]

@nehill197 I have an analysis that needs this. If you are able to review and assuming no big issues, merging by the end of next week would be helpful.

[aylapear]

@nadinehussein This is the PR I was mentioning that adds the skew normal code that we need.

[nehill197]

Okay just one thing I think was wrong (the raw residual calculation).
Then I've suggested a bunch of changes/additions to the tests, and one question about the news file update and how it'll work with fledge.

[joethorley]

@nehill197 - I've accepted all your changes except changing our NEWS files as our fledge system is still in development.

[nehill197]

Sorry @joethorley I approved but then when I was going to merge I realized some of my suggestions were unresolved and you hadn't committed them! Can you take another look at these ones that are still open?