(math/R-robustness) Updated 0.95.1 to 0.99.0

CHANGES in robustbase VERSION 0.99-0 (2023-06-12, svn r928):

  NEW FEATURES:

        * In case it finds an _exact fit_, i.e., k > n/2 residuals are
          (practically) zero, 'lmrob.S()' now returns that, 'scale =
          0', and robustness weights 'weights(., "robustness")' which
          are either 0 or 1 quickly and more consistently.

          This is _not yet_ satisfactorily working for the "large n"
          case.

          **** *TODO*: Optionally (but off by default?? for back
          compatibility??), 'lmrob()' initialized by such an exact-fit
          estimator may use a (user-specified or automatic) scale s^ >
          0 to continue providing efficient SM / MM estimates.

        * 'lmrob.control()' gets class '"lmrobCtrl"' which can be
          'print()'ed and 'update()'d.

        * New tuning const 'zero.scale = 1e-10' instead of hardwired
          'EPS_ZERO' in 'lmrob.S()''s C code; additionally the exact
          fit / zero-residuals checking is finally y-scale
          equivariant, using 1/n ||y||_1 = 'mean(abs(y))'.

  BUG FIXES:

        * Fixed 'covMcd()': corrected the consistency correction
          factor for the reweighted estimate: was MCDCons(p,
          sum(w)/n), now is MCDCons(p, 0.975), see Croux and
          Haesbroeck (1999).

          Consequently, often 'covMcd()' and estimates based on it,
          such as 'BYlogreg()' or 'ltsReg()' are slightly changed.

        * registered 'residuals.lmrob.S' _and_ changed 'formals'
          compatible with generic to '(object, ...)'.

        * 'lmrob.S()' in C level 'refine_fast_s()' no longer modifies
          'beta_cand[]' which notably fixes a bug found in exact-fit
          examples previously resulting in 'coef() == 0' "results",
          also for 'lmrob()' which calls 'lmrob.S()' for its default
          initial estimate.

        * 'print(<lmrob.S>)' finally works again as intended (_not_
          printing a largish list); mostly by fixing internal
          'lmrob.control.minimal()'.

        * 'summary.lmrob(obj)' now also gets a 'weights' component
          from 'obj' if that is non-trivial; this may also correct
          'summary()' output in weighted cases.

        * Minor fixes in 'plot.lts()' related to par mfrow.

        * ltsReg() Help file changed - the note strongly advising NOT
          to use LTS regression removed.

  Misc:

        * argument 'mf' e.g., in 'lmrob.fit()' has been deprecated
          since 2017-12-09 and is defunct now.

        * 'lmrob()' tweaks, notably saying more for 'trace.lev >= 3'.

        * replaceed (almost) all 'Calloc()' by 'R_alloc()' in C code;
          e.g., prevents leakage from user interrupt.

(mef)