fly_select() help says 'fewest photos'; the selection is greedy and not always minimal #106

If we do it: the reference page stops promising a minimum the algorithm does not guarantee. If we never do: a reader of ?fly_select keeps reading mode = "minimal" as the provably smallest set.

Problem

R/fly_select.R:3 and :12 say "minimal" picks "the fewest photos" to reach target coverage. The selection is greedy set cover (fly_select_minimal(), R/fly_select.R:160-205), which does not guarantee the minimum. fly#104's code-check round 2 built a counterexample and ran it through fly_select(): a 4 x 2 km AOI, two 2 km square photos that tile it exactly, and one 2.4 km square photo centred on it. It picks all 3 where the 2 alone give 100%. On the bundled data greedy matches a brute-force minimum at every target from 0.5 to 1.0, so the README's example still holds.

The README, DESCRIPTION and pkgdown text were changed to "a small set" in fly PR #105. The roxygen was outside that diff.

Task

• Reword the two roxygen lines so they do not claim the minimum, e.g. "a small set of photos, adding at each step the one that covers the most new ground"
• devtools::document()