ENH: add indexed loops for maximum, minimum, fmax, fmin (#23177)

Continuation of the ufunc.at optimizations:

add indexed loops for maximum, minimum, fmax, fmin ufuncs and a benchmark for maximum.at (performance increased by ~13x)

* BENCH: add np.maximum.at benchmark

* remove 'explain_chain'

* add a seed to `default_rng() (from review)

* MAINT: formatting and change comments, from review

1 files changed, 6 insertions, 2 deletions

diff --git a/benchmarks/benchmarks/bench_ufunc.py b/benchmarks/benchmarks/bench_ufunc.py
index 090f19224..138f2e9a0 100644
--- a/benchmarks/benchmarks/bench_ufunc.py
+++ b/benchmarks/benchmarks/bench_ufunc.py
@@ -36,13 +36,17 @@ class Broadcast(Benchmark):
 
 class At(Benchmark):
     def setup(self):
-        self.vals = np.random.rand(1_000_000)
-        self.idx = np.random.randint(1000, size=1_000_000).astype(np.intp)
+        rng = np.random.default_rng(1)
+        self.vals = rng.random(10_000_000, dtype=np.float64)
+        self.idx = rng.integers(1000, size=10_000_000).astype(np.intp)
         self.res = np.zeros(1000, dtype=self.vals.dtype)
 
     def time_sum_at(self):
         np.add.at(self.res, self.idx, self.vals)
 
+    def time_maximum_at(self):
+        np.maximum.at(self.res, self.idx, self.vals)
+
 
 class UFunc(Benchmark):
     params = [ufuncs]