Hardcode benchmarks because CI is too unstable

Author: Tortar

docs/src/perf_tips.md

@@ -25,13 +25,18 @@ iter = 1:10^7;
 
 Running with both version we get
 
-```@example 1
-@benchmark fit_iter!(rs, $iter) setup=(rs = ReservoirSampler{Int}(10, AlgRSWRSKIP(); mutable = true))
+```julia
+@btime fit_iter!(rs, $iter) setup=(rs = ReservoirSampler{Int}(10, AlgRSWRSKIP(); mutable = true))
+```
+```julia
+  6.535 ms (2 allocations: 144 bytes)
+```
 ```
-
-```@example 1
 @benchmark fit_iter!(rs, $iter) setup=(rs = ReservoirSampler{Int}(10, AlgRSWRSKIP(); mutable = false))
 ```
+```julia
+  4.816 ms (2 allocations: 144 bytes)
+```
 
 As you can see, the immutable version is 50% faster than 
 the mutable one. In general, more the ratio between reservoir