Add a dedicated S3 class to gs_design_npe() output

[keaven]

Motivation

gsDesign2::gs_design_npe() currently returns an object whose classes are only:

c("tbl_df", "tbl", "data.frame")

The column schema is distinctive and useful (analysis, bound, z,
probability, probability0, theta, info_frac, info, info0, and
info1), but downstream packages cannot identify the object reliably without
matching its column names. In contrast, the other gsDesign2 design
constructors return objects with a package-specific class.

This arose while adding direct support for gs_design_npe() results to
gsDesignBroom::compare_designs():
https://github.com/keaven/gsDesignBroom/issues/2

Suggested enhancement

Could gs_design_npe() add a dedicated S3 class while retaining its existing
tibble classes and behavior? For example:

class(result) <- c("gs_design_npe", class(result))

This would allow downstream packages to register methods or dispatch to a
schema adapter using inherits(x, "gs_design_npe"), rather than relying on
heuristic column-signature detection.

Possible acceptance criteria

• inherits(gs_design_npe(...), "gs_design_npe") is TRUE.
• The result continues to inherit from tbl_df, tbl, and data.frame.
• Existing printing, subsetting, and column names remain unchanged.
• The class is documented as a stable way for downstream packages to identify
  this result type.

This is not blocking the downstream implementation, but it would make the
integration more robust and easier to maintain.

[jdblischak]

Alternatively, if possible, we could restructure the output of gs_design_npe() to match the pattern of the other gs_design_X() functions. Last year I refactored the classes so that the method (eg "ahr", "rd", etc) was stored in the list element design, and the S3 class assigned was either "gs_design" or "fixed_design" (#584). I couldn't easily do this with gs_design_npe() because it only returns the analysis table:

gsDesign2/R/gs_design_npe.R

Lines 444 to 447 in 25cea75

	ans <- ans[order(ans$analysis, ans$bound != "upper"), c("analysis", "bound", "z", "probability", "probability0", "theta", "info_frac", "info", "info0", "info1")]
	rownames(ans) <- NULL
	return(tibble::as_tibble(ans))

whereas the other functions return this table as the list element analysis:

gsDesign2/R/gs_design_ahr.R

Lines 436 to 450 in 25cea75

	ans <- structure(
	list(
	design = "ahr",
	input = input,
	enroll_rate = enroll_rate_out,
	fail_rate = fail_rate,
	bound = bound[is.finite(bound$z), ],
	analysis = analysis
	),
	class = "gs_design",
	binding = binding,
	uninteger_is_from = "gs_design_ahr"
	)
	return(ans)