Fix error/warning forming Tsparse mat w/ Matrix {<,>=}v1.3-0

Context:

In the `Matrix` package v1.3-0, `sparseMatrix` was
[updated](https://cran.r-project.org/web/packages/Matrix/news.html) with a new
argument `repr` and deprecation of `giveCsparse=FALSE`.  Commit
9f8fa00 avoided deprecation warnings when using
`Matrix` package versions >= 1.3-0, but generated errors with versions < 1.3-0.

Action:

Avoid both errors for earlier versions and deprecation warnings for later
versions by constructing Tsparse matrices by directly constructing an
appropriate S4 object.

R-package/quantgen/R/quantile_ensemble.R

@@ -248,9 +248,16 @@ quantile_ensemble_flex = function(qarr, y, tau, weights, tau_groups,
                                   noncross=TRUE, q0=NULL,
                                   lp_solver=c("glpk","gurobi"), time_limit=NULL,
                                   params=list(), verbose=FALSE) {
+
+  # Check inputs
+
+  if (anyDuplicated(tau) != 0L || is.unsorted(tau) || any(tau < 0 | tau > 1)) {
+    stop ("Entries of `tau` must be distinct, sorted in increasing order, >=0, and <=1")
+  }
+
   if (intercept && noncross) {
     stop ("using multiple tau groups with intercept=TRUE and noncross=TRUE is currently unsupported")
-    ## (Matrix formation needs to be debugged or verified for this case.)
+    # (Matrix formation needs to be debugged or verified for this case.)
   }
 
   # Set up some basic objects that we will need
@@ -430,12 +437,11 @@ quantile_ensemble_flex = function(qarr, y, tau, weights, tau_groups,
   }
 
   # Build model$A from parts:
-  model$A = sparseMatrix(
-    i = do.call(c, A_i_parts[seq_len(A_part_ind)]),
-    j = do.call(c, A_j_parts[seq_len(A_part_ind)]),
-    x = do.call(c, A_x_parts[seq_len(A_part_ind)]),
-    dims = c(A_nrow, A_ncol),
-    repr = "T" # faster than "C"; solvers might expect Tsparse form
+  model$A = new("dgTMatrix", # `d`ouble-type entries, `g`eneral structure, `T`sparse (ijx format)
+    i = as.integer(do.call(c, A_i_parts[seq_len(A_part_ind)])) - 1L,
+    j = as.integer(do.call(c, A_j_parts[seq_len(A_part_ind)])) - 1L,
+    x = as.numeric(do.call(c, A_x_parts[seq_len(A_part_ind)])),
+    Dim = as.integer(c(A_nrow, A_ncol))
   )
 
   # Call Gurobi's LP solver, store results