Details of BigDecimal64Array and BigDecimal128Array

[littledan]

BigInt established a pattern of working with TypedArrays and DataViews, to provide serialization for a fixed-size subset. This could be very useful for BigDecimal, where IEEE 754 decimal can provide a good interchange format. Both 64-bit and 128-bit seem to be in use, so I'd propose providing both.

There are two binary formats for IEEE 754 decimal (BID and DPD). Some platforms have hardware acceleration for one, or highly optimized libraries for the other, so I'd suggest we leave it implementation-determined what binary format is the "default", just like endianness in TypedArrays. We could have DataView methods which let you select one or the other.

Any thoughts about this outline? We could leave these arrays for "later" if we want to avoid touching IEEE 754 decimal, but it seems pretty useful to me.