Fix tpca() when there are more samples than features (n > p)

[kimanh-lecao]

Builds on #12; only the last commit is new.

tpca() fails whenever the number of samples exceeds the number of features:

set.seed(1)
tpca(array(rnorm(20 * 20 * 5), dim = c(20, 20, 5)), ncomp = 2)   # n = p: runs
tpca(array(rnorm(21 * 20 * 5), dim = c(21, 20, 5)), ncomp = 2)   # n > p
#> Error in .binary_facewise(a, b, bpparam = NULL) : pa == nb is not TRUE

This is a common shape for targeted panels, for example a few dozen metabolites or antibody titres
measured on more subjects.

Cause. tsvdm() returns the economy-size decomposition: uhat is n x k x t with
k = min(n, p). tpca() then builds the singular values tensor as n x p x t for
uhat %fp% S. When n <= p, k = n and the product conforms; when n > p, uhat has only p columns
against n rows of S, and .binary_facewise() stops.

Fix. Build the singular values tensor with k rows: dim = c(k, p, t) instead of c(n, p, t).
For n <= p this is the same call as before.

Checks.

• n <= p is unchanged: 17 tpca() fits (n < p and n = p; matrix and tensor output; with and
  without centring; several ncomp) are identical() before and after, as are tpls(), tplsda(),
  block_tpls() and block_tplsda() results.
• n > p now runs, and the projections are correct: for t = 5 and 6, with n x p of 21 x 20, 30 x 4,
  46 x 26 and 90 x 31, the tensor variates equal the transformed, centred data times vhat face by
  face to within 2e-14.
• New test in test-tens.tpca.R, mirroring "tpca agrees with mixOmics pca" with n = 10 > p = 3:
  loadings and variates match mixOmics::pca(). It fails on the current main and passes with the
  fix.
• The full test suite passes.

Closes #3.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.