Calculate division and multiplication with Ruby's Integer for large precision

[tompng]

Ruby's Integer implements faster algorithm (karatsuba, toom3, GMP)
Using Integer's div/mult needs base conversion, but it's still faster when precision is large.

# Mult thresholdprec=150;x=BigDecimal('7'*9*prec);10000.times{x*x}# processing time: 0.867585sprec=149;x=BigDecimal('7'*9*prec);10000.times{x*x}# processing time: 1.041640s# Div threasholdprec=299;y=BigDecimal('7'*9*prec);10000.times{BigDecimal(1).div(y,9*prec)}# processing time: 1.426509sprec=300;y=BigDecimal('7'*9*prec);10000.times{BigDecimal(1).div(y,9*prec)}# processing time: 1.299771s

# Performance degrade case.x=BigDecimal('1'*9*300);(x*7).div(x,9*290)#=> 0.7e1# processing time: 0.016612msx=BigDecimal('1'*9*300);(x*7).div(x,9*300)#=> 0.7e1# processing time: 0.209515ms

[tompng]

Closing in favor of #407

Cost of base conversion (not multiplication itself) is quite large for huge n_digits numbers
Example benchmarks:

x=BigDecimal('1'*(10**7));x*x# pull #407: 0.790393s# this patch: 3.530391s# Maximum multiplication supported in #407x=BigDecimal('1'*(9*2**26-1));x*x# pull #407: 28.409065s# this patch: 374.202687sx=BigDecimal('1'*(9*2**24-1));y=BigDecimal('7'*(9*2**23));x.modulo(y)# pull #407: 26.482367s# this patch: 74.415792s

Perhaps this can be used as a fallback of multiplying n_digits > 9*2**26 numbers