Sufficiently big BigUint multiplications fail during CTFE

[just-harry]

When both operands of a std.internal.math.biguintcore.BigUint multiplication exceed a certain threshold (the threshold appears to be 1 << 32) during CTFE, the multiplication will fail as std.internal.math.biguintcore.mulInternal attempts to use core.cpuid.dataCaches to optimise algorithm selection.

This happens only if the operands differ, as squaring follows a different code-path.

Test case:

import std.internal.math.biguintcore : BigUint;
enum largeMultiplication = BigUint.mul(BigUint(ulong(1)) << ulong(33), BigUint(ulong(1)) << ulong(32));

resulting in:

import\core\cpuid.d(124): Error: static variable `_dataCaches` cannot be read at compile time
    const(CacheInfo)[5] dataCaches() { return _dataCaches; }
                                              ^
import\std\internal\math\biguintcore.d(239):        called from here: `dataCaches()`
    return dataCaches[0].size * 1024 / 2;
           ^
import\std\internal\math\biguintcore.d(1715):        called from here: `getCacheLimit()`
        immutable CACHELIMIT = getCacheLimit;
                               ^
import\std\internal\math\biguintcore.d(915):        called from here: `mulInternal(result, cast(const(uint)[])x.data, cast(const(uint)[])y.data)`
            else mulInternal(result, x.data, y.data);
                            ^
example.d(2):        called from here: `mul(BigUint([0u]).this(1LU).opBinary(33LU), BigUint([0u]).this(1LU).opBinary(32LU))`
enum largeMultiplication = BigUint.mul(BigUint(ulong(1)) << ulong(33), BigUint(ulong(1)) << ulong(32));