perlperf: make the example dereference case more reasonable

[tonycoz]

update the results and make the comparison a bit easier to read.

Fixes #10346

I compared a few others:

macos system perl (5.34.1):
Benchmark: timing 10000000 iterations of dereference, direct...
dereference:  0 wallclock secs ( 0.78 usr +  0.00 sys =  0.78 CPU) @ 12820512.82/s (n=10000000)
    direct:  2 wallclock secs ( 0.72 usr +  0.00 sys =  0.72 CPU) @ 13888888.89/s (n=10000000)

macos blead
Benchmark: timing 10000000 iterations of dereference, direct...
dereference:  0 wallclock secs ( 0.71 usr +  0.01 sys =  0.72 CPU) @ 13888888.89/s (n=10000000)
    direct:  1 wallclock secs ( 0.65 usr +  0.00 sys =  0.65 CPU) @ 15384615.38/s (n=10000000)

5.18.2 (x86_64/gcc 4.72:
Benchmark: timing 10000000 iterations of dereference, direct...
dereference:  2 wallclock secs ( 1.93 usr +  0.00 sys =  1.93 CPU) @ 5181347.15/s (n=10000000)
    direct:  3 wallclock secs ( 1.86 usr +  0.00 sys =  1.86 CPU) @ 5376344.09/s (n=10000000)

blead (x86_64/gcc 14)
Benchmark: timing 10000000 iterations of dereference, direct...
dereference:  0 wallclock secs ( 1.14 usr +  0.00 sys =  1.14 CPU) @ 8771929.82/s (n=10000000)
    direct:  0 wallclock secs ( 0.94 usr +  0.00 sys =  0.94 CPU) @ 10638297.87/s (n=10000000)

---

• This set of changes does not require a perldelta entry.

[richardleach]

LGTM