Add separated tests for three string formatting methods

[97littleleaf11]

The three microbenchmarks share the same logic. Currently there are only limited cases, including %s, %d and %f, whose distribution is not exactly the same compared with real world situation.

I would first mark this PR as draft for further discussion.

[97littleleaf11]

% operator:

interpreted: 0.401913s (avg of 5 iterations; stdev 0.54%)
compiled:    0.303943s (avg of 5 iterations; stdev 0.76%)

compiled is 1.322x faster

str.format:

running str_format_format_method
..........
interpreted: 0.581258s (avg of 5 iterations; stdev 0.28%)
compiled:    0.495334s (avg of 5 iterations; stdev 0.66%)

compiled is 1.173x faster

f-string:

running str_format_format_fstring
..........
interpreted: 0.288576s (avg of 5 iterations; stdev 0.83%)
compiled:    0.759397s (avg of 5 iterations; stdev 0.37%)

compiled is 0.380x faster

[97littleleaf11]

after python/mypy#10697:

running str_format_format_method
..........
interpreted: 0.587563s (avg of 5 iterations; stdev 0.35%)
compiled:    0.282582s (avg of 5 iterations; stdev 0.32%)

compiled is 2.079x faster

[97littleleaf11]

after python/mypy#10762:

running str_format_format_method
..........
interpreted: 0.574899s (avg of 5 iterations; stdev 0.25%)
compiled:    0.156026s (avg of 5 iterations; stdev 0.63%)

compiled is 3.685x faster

[97littleleaf11]

after python/mypy#10776:

running str_format_fstring
..........
interpreted: 0.295028s (avg of 5 iterations; stdev 2.3%)
compiled:    0.188493s (avg of 5 iterations; stdev 1.7%)

compiled is 1.565x faster