Add package scaffolding and CI

[nickreich]

Sets up an empty but checkable package, so the CI pipeline is validated before
any code depends on it.

This is the first of a stack of six PRs porting
alloscore to a hubverse-native
alloscore2. Each layer is self-contained and passes R CMD check on its own;
Imports grows only as code actually uses each package, so there are no
"declared but unused" NOTEs along the way.

Contents

DESCRIPTION, NAMESPACE, NEWS.md, LICENSE.md, five GitHub Actions
workflows with SHA-pinned actions, .lintr, air.toml, codecov.yml,
_pkgdown.yml, .Rbuildignore and the .github community files, all adapted
from hubEvals. The only R file is the package doc.

What this is meant to prove

• the check matrix runs on all five platforms
• lintr::lint_package() and air format . --check gate as intended
• the pkgdown site builds, and hubStyle installs via Config/Needs/website
• Codecov reporting is wired up

Heavier dependency resolution (arrow, hubUtils, hubExamples from
r-universe) is first exercised in PRs 2 and 3.

Three deliberate deviations from the hubEvals template

• Licence is GPL (>= 3), not MIT. alloscore2 is a derivative work of
  aaronger/alloscore, which is GPL-3, so the licence cannot be changed without
  the original author's consent. Aaron Gerding is recorded in DESCRIPTION as
  aut, cph.
• Docs publish to GitHub Pages, not Netlify. reichlab repositories do not
  use Netlify, so the inherited pkgdown-netlify-preview.yaml is replaced with
  the standard r-lib pkgdown.yaml, matching reichlab/distfromq. It still
  builds the site on every PR and deploys to gh-pages on push to main.
  Actions stay SHA-pinned, which distfromq's copy does not do.
• .lintr relaxes two linters. The allocation literature's notation is part
  of this package's public API — F is a predictive cdf, Q a quantile
  function, K a budget, O and U over- and under-prediction costs, Lambda
  a marginal expected benefit — and eps_K and qs_OK are documented argument
  and column names inherited from the original package. object_name_linter and
  T_and_F_symbol_linter flag all of those, so both are relaxed rather than
  renaming public API. indentation_linter is disabled because it disagrees
  with Air, which owns formatting here. The reasoning is in a comment at the top
  of the file.

Verified locally

R CMD check: Status: OK. lintr::lint_package(): 0 lints.
air format . --check: clean.

[nickreich]

remove this author from the list.

[nickreich]

Closed and locked by GitHub after a force-push to an unrelated history (my error: the branch was regenerated on a recreated main with the same tree but a different root commit). Replaced by #8, which uses the same branch with the reviewed content plus the requested DESCRIPTION change.