Add show/coeftable/confint to models; KaplanMeier show (closes #78, #79)

[andreasnoack]

Summary

Gives every model class a terse, faithful show plus an explicit coeftable/confint inference layer, and removes the non-idiomatic summary(::Cox). Closes #78 and #79.

The design follows the S/R print-vs-summary convention: show renders only stored state (coefficients, baseline, fit statistics) and never computes covariance-derived quantities; inference is requested explicitly via coeftable/confint, which take a level keyword that a zero-argument display can't. coeftable returns a StatsBase.CoefTable (the ecosystem-standard, matching GLM.jl/MixedModels); confint returns a DataFrame.

Cox

• Add coef / coefnames / vcov; coeftable (with exp(coef) and its confidence interval on the hazard-ratio scale, matching R's summary(coxph)) and confint.
• Terse show (method, n, events, loglik/AIC/BIC, coefficients). The O(n²) C-index is no longer computed in show — it stays available via harrells_c.
• Remove summary(::Cox) (Remove summary(::Cox): it overloads Base.summary with non-idiomatic (R-style) semantics #79): it overloaded Base.summary, whose contract is a one-line descriptor, with a full R-style table.

GeneralHazardModel (#78)

Previously had no show at all — it fell back to Julia's default struct dump (every data matrix printed).

• Add coefnames, coeftable, confint, and a terse show.
• coeftable/confint cover the covariate effects only; baseline distribution parameters are shown by show as the fitted distribution, not tested — a null of zero is the question for a covariate effect but not for a baseline shape/scale.
• The GH model's two coefficient sets are grouped and labelled by role (hazard-level vs time-scale), since a covariate can enter both; PH/AFT/AH expose their single active block.

KaplanMeier

• Add a show (n, events, median survival) replacing the struct dump.
• fit now errors on right-hand-side variables — KaplanMeier is a single-curve estimator, and silently returning a pooled curve for ~ g is a footgun (R's survfit stratifies). Stratification is tracked in [KaplanMeier] Support stratified estimation via formula RHS (grouped survival curves) #83; this error is the interim guardrail it refers to.

Shared / tests / docs

• _coef_fmt, _ci_levstr, _hr_coeftable factored into utils.jl so Cox and the parametric models build identical tables.
• Reference-string tests via repr(MIME"text/plain"(), m) for each model class (deterministic direct-construction / StableRNG fits), plus coeftable/confint structure and relationship checks.
• Docs updated: cox.md (summary→coeftable/confint), fitting.md (GHM coeftable/confint), example.md (each parametric class now displays its show).

734 tests pass; docs build locally.

[codecov]

Codecov Report

❌ Patch coverage is 82.14286% with 20 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.12%. Comparing base (086fb1e) to head (68a51ba).

Files with missing lines	Patch %	Lines
src/Parametric/GeneralHazard.jl	72.60%	20 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #84      +/-   ##
==========================================
- Coverage   92.32%   92.12%   -0.20%     
==========================================
  Files          12       12              
  Lines         873      978     +105     
==========================================
+ Hits          806      901      +95     
- Misses         67       77      +10     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.