- Improve the accuracy of fast pass' range reduction.
- Provide tighter error estimations.
- Reduce the table size when `LIBC_MATH_SMALL_TABLES` flag is set.

Fix #98393

The following errors were reported by Paul Zimmerman:
```
expm1
Running worst cases check in --rndn mode...
FAIL x=0x1.c3b93d3b25c91p+8 ref=0x1.9fb21849affd1p+651 z=0x1.9f921849affd1p+651
Running worst cases check in --rndz mode...
FAIL x=0x1.d6479eba7c971p+8 ref=0x1.62a88613629b5p+678 z=0x1.62a886135e9b5p+678
Running worst cases check in --rndu mode...
FAIL x=0x1.1f0da93354198p+7 ref=0x1.0bd73b73fc74dp+207 z=0x1.0bd73b73fc74cp+207
Running worst cases check in --rndd mode...
FAIL x=0x1.93b3037166891p+8 ref=0x1.554f3a5c4535cp+582 z=0x1.554f3a5c4535bp+582
```
```
tan
Running worst cases check in --rndn mode...
FAIL x=0x1.00452f0e0134dp-13 ref=0x1.00452f2367da9p-13 z=0x1.00452f2367daap-13
Running worst cases check in --rndz mode...
FAIL x=0x1.89f0f5241255bp-2 ref=0x1.9e997dd861117p-2 z=0x1.9e997dd861118p-2
FAIL x=0x1.045457ae3994p+5 ref=0x1.0c06941768ef3p+1 z=0x1.0c06941768ef4p+1
```

Using the same range reduction as `sin`, `cos`, and `sincos`:
1) Reducing `x = k*pi/128 + u`, with `|u| <= pi/256`, and `u` is in
double-double.
2) Approximate `tan(u)` using degree-9 Taylor polynomial.
3) Compute
```
   tan(x) ~ (sin(k*pi/128) + tan(u) * cos(k*pi/128)) / (cos(k*pi/128) - tan(u) * sin(k*pi/128))
```
using the fast double-double division algorithm in [the CORE-MATH
project](https://gitlab.inria.fr/core-math/core-math/-/blob/master/src/binary64/tan/tan.c#L1855).
4) Perform relative-error Ziv's accuracy test
5) If the accuracy tests failed, we redo the computations using 128-bit
precision `DyadicFloat`.

Fixes https://github.com/llvm/llvm-project/issues/96930

Context: https://github.com/llvm/llvm-project/pull/87017

- Add proxy header `libc/hdr/math_macros.h` that will:
  - include `<math.h>` in overlay mode,
- include `"include/llvm-libc-macros/math-macros.h"` in full build mode.
- Its corresponding CMake target `libc.hdr.math_macros` will only depend
on `libc.include.math` and `libc.include.llvm-libc-macros.math_macros`
in full build mode.
- Replace all `#include "include/llvm-libc-macros/math-macros.h"` with
`#include "hdr/math_macros.h"`.
- Add dependency to `libc.hdr.math_macros` CMake target when using
`add_fp_unittest`.
- Update the remaining dependency.
- Update bazel overlay: add `libc:hdr_math_macros` target, and replacing
all dependency on `libc:llvm_libc_macros_math_macros` with
`libc:hdr_math_macros`.

Reland of #84991

A downstream overlay mode user ran into issues with the isnan macro not
working in our sources with a specific libc configuration. This patch
replaces the last direct includes of math.h with our internal
math_macros.h, along with the necessary build system changes.

The semantics for casting can range from "bitcast" (same representation)
to "different representation", to "type promotion". Here we remove the
cast operator and force usage of `get_val` as the only function to get
the floating point value, making the intent clearer and more consistent.

`StorageType` may be a `BigInt` under the hood. Initializing it with
`-1` does not yields the maximum value.

This one might be a bit controversial since the terminology has been
introduced from the start but I think `FRACTION_LEN` is a better name
here. AFAICT it really is "the number of bits after the decimal dot when
the number is in normal form."

`MANTISSA_WIDTH` is less precise as it's unclear whether we take the
leading bit into account.
This patch also renames most of the properties to use the `_LEN` suffix
and fixes useless casts or variables.

Clean up usage of `DECLARE_SPECIAL_CONSTANTS` in global scope.