[benchmark] Fix quantile estimation type

The correct quantile estimation type for printing all measurements in the summary report while `quantile == num-samples - 1` is R-1, SAS-3.  It's the inverse of empirical distribution function.

References:
* https://en.wikipedia.org/wiki/Quantile#Estimating_quantiles_from_a_sample
* discussion in https://github.com/apple/swift/pull/19097#issuecomment-421238197

benchmark/scripts/compare_perf_tests.py

@@ -34,8 +34,7 @@ import re
 import sys
 from bisect import bisect, bisect_left, bisect_right
 from collections import namedtuple
-from decimal import Decimal, ROUND_HALF_EVEN
-from math import sqrt
+from math import ceil, sqrt
 
 
 class Sample(namedtuple('Sample', 'i num_iters runtime')):
@@ -143,15 +142,12 @@ class PerformanceTestSamples(object):
         return self.samples[-1].runtime
 
     def quantile(self, q):
-        """Return runtime of a sample nearest to the quantile.
+        """Return runtime for given quantile.
 
-        Explicitly uses round-half-to-even rounding algorithm to match the
-        behavior of numpy's quantile(interpolation='nearest') and quantile
-        estimate type R-3, SAS-2. See:
+        Equivalent to quantile estimate type R-1, SAS-3. See:
         https://en.wikipedia.org/wiki/Quantile#Estimating_quantiles_from_a_sample
         """
-        index = int(Decimal((self.count - 1) * Decimal(q))
-                    .quantize(0, ROUND_HALF_EVEN))
+        index = max(0, int(ceil(self.count * float(q))) - 1)
         return self.samples[index].runtime
 
     @property