make binary is not deterministic

I've been tracking a bunch of non-determinism issues we have in our build and a pretty fundamental one seems to be that the make binary produced by `rules_foreign_cc` is not deterministic, yielding to downstream invalidations of anything using `rules_foreign_cc`.

Minimal repro:

```
$ bazel clean
$ bazel build --config=no-remote @rules_foreign_cc//toolchains:make_tool && cp bazel-bin/external/rules_foreign_cc/toolchains/make/bin/make make.1
$ bazel clean
$ bazel build --config=no-remote @rules_foreign_cc//toolchains:make_tool && cp bazel-bin/external/rules_foreign_cc/toolchains/make/bin/make make.2
$ md5sum make.1 make.2
```

After a lot of pain diagnosing other proximate issues that weren't the root cause, I was able to fix this by patching `foreign_cc/built_tools/make_build.bzl` to use `make install-strip` instead of `make install`. Assuming `ar` is correctly configured in the toolchain to use `-D`, then this seems to be sufficient.

I also wonder if the default targets passed to `make` and `configure_make` should default to `install-strip` (maybe depending on C compiler settings)? This seems hard in the general case because there is no guarantee that the built packages follow the Automake conventions... but I wonder if something could be done to make this issue more prominent, in the docs or elsewhere, as otherwise it seems too easy to trigger this footgun.

Start with foreign_cc/built_tools/make_build.bzl and reproduce the two builds from the issue using bazel clean, the no-remote configuration, and md5sum. Check whether the produced make binaries match across builds; done means the reported non-determinism is removed, while the broader defaults and documentation question is resolved or explicitly scoped.

Written by the indexing model from the issue text.