context: implement Context.Value using iteration rather than recursion

context: implement Context.Value using iteration rather than recursion

In profiles of a production server, 2.3% of CPU time is spent in
runtime.newstack for stacks with 12 or more chained Context.Value
calls. Using iteration will avoid some needless stack resizes.

When calling Context.Value in the same goroutine (see
DeepValueSameGoRoutine) , no stack resizing is needed (after warming
up), and this change reduces time/op by around 10%.

The time to start a new goroutine and call Context.Value (see
DeepValueNewGoRoutine) is reduced by over 75% if a stack resize is
needed. If you factor out the overhead of starting the goroutine
(about 960ns) then, avoiding the stack resize saves about 95%.

```
name                                          old time/op  new time/op  delta
CommonParentCancel-12                          960ns ± 1%   958ns ± 1%     ~     (p=0.561 n=9+9)
WithTimeout/concurrency=40-12                 1.31µs ± 2%  1.29µs ± 6%     ~     (p=0.305 n=9+10)
WithTimeout/concurrency=4000-12               1.30µs ± 2%  1.30µs ± 2%     ~     (p=0.343 n=10+10)
WithTimeout/concurrency=400000-12             1.03µs ± 1%  1.02µs ± 2%     ~     (p=0.213 n=9+9)
CancelTree/depth=1/Root=Background-12          123ns ± 5%   126ns ± 2%   +2.61%  (p=0.023 n=10+9)
CancelTree/depth=1/Root=OpenCanceler-12        781ns ± 4%   806ns ± 4%   +3.20%  (p=0.022 n=9+10)
CancelTree/depth=1/Root=ClosedCanceler-12      370ns ± 4%   369ns ± 3%     ~     (p=0.497 n=9+10)
CancelTree/depth=10/Root=Background-12        4.74µs ± 4%  4.78µs ± 3%     ~     (p=0.516 n=10+10)
CancelTree/depth=10/Root=OpenCanceler-12      6.31µs ± 4%  6.29µs ± 4%     ~     (p=1.000 n=10+10)
CancelTree/depth=10/Root=ClosedCanceler-12    2.10µs ± 5%  2.09µs ± 5%     ~     (p=0.839 n=10+10)
CancelTree/depth=100/Root=Background-12       51.0µs ± 3%  51.2µs ± 2%     ~     (p=0.631 n=10+10)
CancelTree/depth=100/Root=OpenCanceler-12     60.8µs ± 1%  61.6µs ± 4%     ~     (p=0.274 n=8+10)
CancelTree/depth=100/Root=ClosedCanceler-12   19.3µs ± 2%  19.0µs ± 3%     ~     (p=0.123 n=10+10)
CancelTree/depth=1000/Root=Background-12       504µs ± 4%   512µs ± 4%     ~     (p=0.123 n=10+10)
CancelTree/depth=1000/Root=OpenCanceler-12     615µs ± 6%   619µs ± 4%     ~     (p=1.000 n=10+10)
CancelTree/depth=1000/Root=ClosedCanceler-12   190µs ± 2%   192µs ± 3%     ~     (p=0.190 n=9+9)
CheckCanceled/Err-12                          12.1ns ± 2%  12.1ns ± 2%     ~     (p=0.615 n=10+10)
CheckCanceled/Done-12                         7.27ns ± 1%  7.26ns ± 1%     ~     (p=0.698 n=10+10)
ContextCancelDone-12                          1.03ns ± 1%  1.03ns ± 1%     ~     (p=0.474 n=9+9)
DeepValueNewGoRoutine/depth=10-12             1.02µs ± 3%  0.99µs ± 2%   -3.41%  (p=0.000 n=10+10)
DeepValueNewGoRoutine/depth=20-12             1.11µs ± 3%  1.08µs ± 2%   -2.51%  (p=0.004 n=10+10)
DeepValueNewGoRoutine/depth=30-12             5.55µs ±10%  1.17µs ± 4%  -78.91%  (p=0.000 n=10+10)
DeepValueNewGoRoutine/depth=50-12             5.70µs ±13%  1.35µs ± 2%  -76.31%  (p=0.000 n=10+10)
DeepValueNewGoRoutine/depth=100-12            9.69µs ± 4%  1.82µs ± 2%  -81.18%  (p=0.000 n=10+10)
DeepValueSameGoRoutine/depth=10-12            54.2ns ± 2%  46.8ns ± 2%  -13.71%  (p=0.000 n=9+9)
DeepValueSameGoRoutine/depth=20-12             109ns ± 2%    97ns ± 2%  -11.11%  (p=0.000 n=10+10)
DeepValueSameGoRoutine/depth=30-12             155ns ± 3%   140ns ± 1%   -9.49%  (p=0.000 n=10+10)
DeepValueSameGoRoutine/depth=50-12             256ns ± 2%   226ns ± 2%  -11.83%  (p=0.000 n=10+10)
DeepValueSameGoRoutine/depth=100-12            492ns ± 3%   442ns ± 1%  -10.15%  (p=0.000 n=10+10)
```

Fixes #47292

Change-Id: I6bdeb234c979fb8fd6bfb91fd345cb5038f52c75
Reviewed-on: https://go-review.googlesource.com/c/go/+/335790
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Damien Neil <dneil@google.com>