optimize std.base64

Remove the Unsafe varient, as it has no advantages.

Also remove the Ignore varient, because while it might be necessary
to implement S/MIME it currently has no users, and I think it would
be better implemented with a streaming API. (The blocks are only
4 long). At that time we could add a decodeBlock() function.

About a 80% speed up of Base64Decode.

opt-encode: 1008.9MiB/sec
opt-decode: 711.7MiB/sec
noopt-encode: 1.4GiB/sec
noopt-decode: 557.8MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 882.4MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 548.2MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 903.3MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 552.7MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 903.6MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 554.2MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 892.6MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 553.8MiB/sec
opt-encode: 1.4GiB/sec
opt-decode: 892.7MiB/sec
noopt-encode: 1.4GiB/sec
noopt-decode: 551.3MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 905.5MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 555.5MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 904.0MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 553.9MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 903.6MiB/sec
noopt-encode: 1.4GiB/sec
noopt-decode: 550.5MiB/sec
opt-encode: 1.5GiB/sec
opt-decode: 891.5MiB/sec
noopt-encode: 1.5GiB/sec
noopt-decode: 550.0MiB/sec