#4955 Numeric clauses admit argument types their callee cannot handle

[David Scherfgen]

Ordinary spellings of erfc, lambert_w, elliptic_eu, beta_incomplete_regularized, elliptic_pi and inverse_erfc crash with raw Lisp type errors

Maxima 5.49post, SBCL 2.6.7.

Each of these is a well-posed call whose numeric clause accepts the argument and then hands it to a routine that cannot take it.

(%i1) display2d:false$
(%i2) print(errcatch(ev(erfc(1+%i),numer)))$
The value 1 is not of type FLOAT
[]
(%i3) print(errcatch(ev(lambert_w(1+%i),numer)))$
There is no applicable method for the generic function
  #<STANDARD-GENERIC-FUNCTION BIGFLOAT-IMPL:EPSILON (4)>
when called with arguments (1).
[]
(%i4) print(errcatch(elliptic_eu(1/2+%i/3, 0.5)))$
The value ((RAT SIMP) 1 2) is not of type REAL
[]
(%i5) print(errcatch(beta_incomplete_regularized(-3b0,2b0,1/2)))$
The value ((BIGFLOAT SIMP 56) -54043195528445952 2) is not of type NUMBER
[]
(%i6) print(errcatch(elliptic_pi(-3,-2.0,1.5)))$
The value ((MPLUS SIMP) 0.8906829430383684 ((MTIMES SIMP) -0.5906532424614204 $%I))
is not of type NUMBER
[]
(%i7) print(errcatch(inverse_erfc(0.5+0.5*%i)))$
There is no applicable method for the generic function
  #<STANDARD-GENERIC-FUNCTION BIGFLOAT-IMPL:FLOAT (3)>
when called with arguments (#C(0.5 -0.5) 0.0).
[]

Expected: a value, or a Maxima-level error, not a Lisp type error. Note elliptic_pi(-3,2.0,1.5) with POSITIVE phi works and returns 0.8906829430383684-0.5906532424614204*%i; only the negative-phi case, where the result is complex, crashes.