Second-order plotting: draw bin averages at geometric bin centres (#382)

[gustavdelius]

Part of #377 — the rendering counterpart of the bin-averaging programme. Once a
quantity is a bin average, plotting it at the bin boundary is wrong.

The problem

All size-resolved plots route their x-locations through get_ArraySpeciesBySize_w(),
which returns the grid nodes params@w / params@w_full — the left bin
edges. A bin average N_j = (1/Δw_j)∫_{w_j}^{w_{j+1}} N dw does not live at
the boundary, so it is mis-placed by half a bin.

Where averaged values belong

The geometric bin centre w*_j = sqrt(w_j w_{j+1}) = w_j·sqrt(β) — log-symmetric,
second-order-correct, and exact for the community spectrum N ∝ w^{-2}. A
uniform half-bin shift to the right on the log axis.

Design

A representation tag ("point" / "average", default "point") on the
ArraySpeciesBySize and ArrayTimeBySpeciesBySize constructors:

• Producers tag honestly. The bin-averaged sinks getPredMort(),
  getFMort(), getMort(), getExtMort() and the reproductive investment
  getERepro() tag themselves "average"; point-valued quantities (encounter,
  growth) keep the "point" default.
• One placement point, gated on the flag. get_ArraySpeciesBySize_w() and
  the new get_ArrayTimeBySpeciesBySize_w() return the geometric bin centres
  (new internal bin_midpoints() helper) for "average" arrays only when
  second_order_w[["bin_average"]] is set. So first-order/default models, their
  as.data.frame() output and all vdiffr snapshots are unchanged; a
  second-order model is internally consistent across dynamics, diagnostics and
  plot placement.
• The tag is preserved through the [ subset, the time-slice, and the
  MizerSim rate path (sim_size_rate / get_species_size_rate_from_sim thread a
  representation argument to the constructors). Everything downstream
  (prepare_*_plot_data, as.data.frame, plot2) follows the central helpers
  automatically.

Scope

plotSpectra/plotCDF build their own data frames and apply the w^power
weight; their centre placement (and the CDF's opposite, edge placement) is the
separate #383, which reuses bin_midpoints().

Tests

test-ArraySpeciesBySize.R / test-ArrayTimeBySpeciesBySize.R: producers tag
honestly; bin_midpoints are the geometric centres; the size axis shifts to the
centres only under second_order_w (point quantities stay on nodes);
as.data.frame x-values follow the tag; subsetting/slicing preserve it. The
rate-function and backwards-compat snapshots gained only the new benign
representation attribute (accepted). Full suite: 2824 pass.

Docs

NEWS entry; a "Plotting follows the same distinction" note in the
"Point values and bin averages" section of the numerical-details vignette.

🤖 Generated with Claude Code