vignettes: explain proportional baseline

[carmen-cabrera]

Summary

• clarifies the proportional baseline narrative in the coverage-bias vignette
• replaces the previous coverage-bias/population plot with a two-panel visual
• shows observed user counts against the constant-coverage baseline and residual departures from that baseline
• adds a NEWS note for the vignette improvement

Checks

• git diff --check
• local pkgdown preview built successfully

Closes #67

[fcorowe]

This is looking good overall, and the new visual addresses #67 nicely. I found one small wording issue to tighten before merge: the prose should distinguish the count residual from the default selected residual returned by validate_bias_residual_structure().

[fcorowe]

Small wording fix before merge: this paragraph says the function returns residuals as observed-minus-expected user counts, but the default selected residual from validate_bias_residual_structure() is coverage_score_residual. Could we rephrase to make both residual concepts explicit?

Suggested replacement:

`debiasR::validate_bias_residual_structure()` calculates this baseline and
returns expected user counts plus several residual measures for each LAD. These
include count residuals, defined as observed user counts minus the counts
expected under constant coverage, and the default `coverage_score_residual`,
defined as local coverage minus overall coverage.