prim/BitFlag: assert bit index in makeMask

[AllyOmega]

Fixes #85.

Change

sead::BitFlag<T>::makeMask gains the missing bit-index assertion:

static T makeMask(int bit)
{
    SEAD_ASSERT(static_cast<u32>(bit) < sizeof(T) * 8);
    return T(1) << bit;
}

Rationale

• Issue prim/BitFlag: Missing assertions in isOnBit and other functions #85 asks for the missing static_cast<u32>(bit) < sizeof(T)*8 assertions in the
  BitFlag bit-index functions (setBit, resetBit, changeBit, toggleBit,
  isOnBit, isOffBit, testAndClearBit).
• The original sead lineage (aboood40091/sead,
  engine/library/include/prim/seadBitFlag.h) ships this assertion inside makeMask
  (currently commented out there), covering all of the above in one place — they are all
  makeMask callers.
• BitFlagUtil::countRightOnBit in the same module already carries the identical
  assertion pattern; this restores consistency for the template.

Verification

• Release codegen is unchanged: a TU instantiating BitFlag32/BitFlag64 and
  exercising all bit-index functions compiles to a byte-identical .o with and without
  this patch (clang 15.0.7, -O3, -fno-exceptions, no SEAD_DEBUG), because
  SEAD_ASSERT collapses to if (false) in release builds.
• With -DSEAD_DEBUG=1 the same TU compiles with the asserts active (assert call sites
  present; TU size grows 1520 -> 2792 bytes).
• clang-format 12.0.1 (the CI-pinned version) passes on the modified header.

Caveat

The issue text ends with "and more" — without the original game's debug symbols for these
functions, further missing assertions cannot be enumerated with confidence. This PR is the
minimal, lineage-backed restoration; more can follow with evidence.

---

This change is 

[german77]

This assert should generate the following pseudo code. Currently just based on the missing cBitNum this implementation is incorrect

    if (0xf < param_3) {
      sead::system::HaltWithDetailNoFormat("...include/prim/seadBitFlag.h",  ...LineNumber... ,"static_cast<u32>(bit) < cBitNum");
    }
    *(ushort *)(this + 0x70) = *(ushort *)(this + 0x70) | (ushort)(1 << (ulong)(param_3 & 0x1f));

[AllyOmega]

Thanks for the pseudo code — that settles it. Updated: BitFlag<T> now exposes static constexpr u32 cBitNum = sizeof(T) * 8; and makeMask asserts static_cast<u32>(bit) < cBitNum, so the assert string matches the binary ("static_cast<u32>(bit) < cBitNum"), and the halt fires for bit > 15 on a 16-bit flag exactly as in your pseudo code (0xf < param_3).

Re-verified after the change: release codegen is still byte-identical to the pre-patch baseline (clang 15.0.7, -O3, no SEAD_DEBUG), the debug build compiles with asserts active, and clang-format 12 passes.

If the cBitNum placement/usage still doesn't match your evidence (e.g. it should live in BitFlagUtil or be an enum), tell me and I'll adjust or close — I don't have the binary you're looking at, so your decompiled line is the ground truth here.