This rework changes two important things

1) the output redirect is deactivated while control is given to the reporters.
   This means that combining reporters that write to stdout with capturing
   reporters, e.g. `./tests -s -r console -r junit::out=junit.xml`, no
   longer leads to the capturing reporter seeing all the output from
   the other reporter captured.

Trying this with the `SelfTest` binary would previously lead to JUnit
spending **hours** trying to escape all of ConsoleReporter's output and
write it to the output file. I actually ended up killing the process
after 3 hours, during which the JUnit reporter wrote something like 50 MBs
of output to a file.

2) The redirect object's lifetime is tied to the `RunContext`, instead
  of being constructed for every partial test case run separately.

This has no effect on the basic StreamRedirect, but improves the FileRedirect
significantly. Previously, running many tests in single process with this
redirect (e.g. running `SelfTest -r junit`) would cause later tests to
always fail before starting, due to exceeding the limit of temporary files.

For the current `SelfTest` binary, the old implementation would lead to
**295** test failures from not being able to initiate the redirect. The
new implementation completely eliminates them.

----

There is one downside to the new implementation of FileRedirect, specific
to Linux. Running the `SelfTest` binary on Linux causes 3-4 tests to have
no captured stdout/stderr, even though the tests were supposed to be
writing there (there was no output to the actual stdout/stderr either,
the output was just completely lost).

Since this never happen for smaller test case sets, nor does it reproduce
on other platforms, this implementation is still strictly better than
the old one, and thus it can get reasonably merged.

Move `catch_case_sensitive.hpp` from directory
`src/catch2/internal/internal` to be directly under `src/catch2/`.

Header `catch_case_sensitive.hpp` defines `enum CaseSensitive` which is
part of the public API, since it is exposed by the string matcher
contracts.

Fixes issue caught by Clang Tidy misc-include-cleaner check.

This is done by no longer requiring all assertions to be seen
by the JUnit reporter. Since the JUnit reporter never outputs
all assertions, even with `-s`, the only difference from storing
passing assertions in the `CumulativeReporterBase` was some
bookkeeping in deciding which sections are relevant to the output.

Given `TEST_CASE` that just runs many passing assertions, e.g.

```
TEST_CASE( "PERFORMANCE_TEST_CASE", "[.]" ) {
    for (size_t i = 0; i < 1000'000'000; ++i) {
        REQUIRE( i == i );
    }
}
```

the new JUnit reporter will finish in 5:47, using up ~7.7 MB of RAM.
The old JUnit reporter would finish in 0:30, due to bad_alloc
after using up 14.5 GB of RAM (the system has 16 GB total).

If the total number of assertions is lowered to 10M, the old
implementation uses up ~7.1 GB of RAM and finishes in 12 minutes.
The new implementation still needs only ~7.7 MB of RAM, and finishes
in 4 minutes.

There is a slight downside in that the output is slightly different;
the new implementation will include the `TEST_CASE` level section
in output, even if it does not have its own assertion. In other words,
given a `TEST_CASE` like this

```
TEST_CASE( "JsonWriter", "[JSON][JsonWriter]" ) {
    std::stringstream stream;
    SECTION( "Newly constructed JsonWriter does nothing" ) {
        Catch::JsonValueWriter writer{ stream };
        REQUIRE( stream.str() == "" );
    }

    SECTION( "Calling writeObject will create an empty pair of braces" ) {
        { auto writer = Catch::JsonValueWriter{ stream }.writeObject(); }
        REQUIRE( stream.str() == "{\n}" );
    }
}
```

the new implementation will output 3 `testcase` tags, 2 for the explicit
 `SECTION`s with tests, and 1 for the top level section.

However, this can be worked-around if required, and the performance
improvement is such that it is worth changing the current output,
even if it needs to be fixed in the future.

Closes #2897