Add sqrt() and rsqrt() to Sk4f.

src/core/Sk4x.h

 #ifndef Sk4x_DEFINED
 #define Sk4x_DEFINED
 
 #include "SkTypes.h"
 
 #define SK4X_PREAMBLE 1
     #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
         #include "Sk4x_sse.h"
     #else
         #include "Sk4x_portable.h"
@@ skipping 32 matching lines @@
 
     Sk4x   bitNot()            const;
     Sk4x   bitAnd(const Sk4x&) const;
     Sk4x    bitOr(const Sk4x&) const;
     // TODO: Sk4x bitAndNot(const Sk4x&) const; is efficient in SSE.
     Sk4x      add(const Sk4x&) const;
     Sk4x subtract(const Sk4x&) const;
     Sk4x multiply(const Sk4x&) const;
     Sk4x   divide(const Sk4x&) const;
 
+    Sk4x rsqrt() const;   // Approximate reciprocal sqrt().
+    Sk4x  sqrt() const;   // this->multiply(this->rsqrt()) may be faster, but less precise.
+
     Sk4i            equal(const Sk4x&) const;
     Sk4i         notEqual(const Sk4x&) const;
     Sk4i         lessThan(const Sk4x&) const;
     Sk4i      greaterThan(const Sk4x&) const;
     Sk4i    lessThanEqual(const Sk4x&) const;
     Sk4i greaterThanEqual(const Sk4x&) const;
 
     static Sk4x Min(const Sk4x& a, const Sk4x& b);
     static Sk4x Max(const Sk4x& a, const Sk4x& b);
 
@@ skipping 21 matching lines @@
 #undef SK4X_PRIVATE
 };
 
 #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
     #include "Sk4x_sse.h"
 #else
     #include "Sk4x_portable.h"
 #endif
 
 #endif//Sk4x_DEFINED