1) In all cases is better have only the routines that you use.
2) External memory don't reduces speed, only in certain cases.
3) Yes

I) 58K is not small is larger program, the max in 51 derivate is 64K. Why not?. If program runs in pentium (processor) at 500MHz is faster than 8051 (microcontroller)at 12 MHz.
II) If you make your oun routines you can know the exact times. But if you are new may be you can't do accurate code. Many 'other routines' tell you the time spent or you can count instructions and the time in each one.
III) Simulators run in PC depends of your PC speed. But I think you want mean EMULATOR. Search in this forum the diferences.