Fix/warnings 2

[Foxpunk]

Problem

• Several switch statements over enum-like fields didn't handle unexpected
  values explicitly, which could lead to silent incorrect behavior.
• Return values of time(), fseek() and snprintf() were ignored,
  even though their failure (or truncation, for snprintf()) could lead
  to incorrect behavior going unnoticed.
• A loop variable used MetaId (USHORT), a type not intended for
  holding counts, to iterate up to m_info.pin_count (type size_t),
  risking narrowing and a potential infinite loop.
• MessageSet used int for message codes that are always within a
  small fixed range, wider than necessary.

Fix

• Added explicit default: fb_assert(false) branches to catch unexpected
  switch values early instead of failing silently.
• Saved the return values of time(), fseek() and snprintf() into
  [[maybe_unused]] variables and added fb_assert checks on them.
• Changed the loop variable type to FB_SIZE_T to match the type of
  m_info.pin_count it's compared against, avoiding narrowing.
• Changed MessageSet fields to USHORT.

Notes

• Changing pin_count's type isn't an option since PerformanceInfo is
  a public struct used outside this file — that would be an API change.
  Casting pin_count at the comparison would just hide the mismatch
  rather than fix it: i is used as an array index into
  m_tableCounters, not as a MetaId, so MetaId was the wrong type
  here regardless of pin_count's type. Widening the loop variable to
  FB_SIZE_T fixes the actual type mismatch and removes the risk of
  overflow/infinite loop if pin_count ever grows beyond USHORT range.
• MessageSet fields were changed to USHORT. All current usages
  (MAIN_USAGE, EXAMPLES) are initialized with message codes in range
  3..42, e.g. MAIN_USAGE{{3, 21}, {41}}, EXAMPLES{{22, 27}, {42}} —
  well within USHORT (0..65535). If new codes are added in the future,
  make sure they still fit into USHORT.

[craftmaster1231]

Write this in PR description, not in code. Keep newlines between declarations.

[craftmaster1231]

Don't use C-style cast. Use static_cast for that.

[craftmaster1231]

Use different name const int variables. resultEnd and resultSet, for example.

[craftmaster1231]

Explain this in PR. Why not to change type or cast m_info.pin_count instead?

[craftmaster1231]

Rename the PR to reflect changes you made. Write PR description.
Don't write about warnings in PR name. Better don't write about them at all. Write about problems and fixes.

[craftmaster1231]

One of your commit messages: use FB_SIZE_T instead of MetaId for loop variable in TraceRuntimeStats (bugprone-too-small-loop-variable)

Please remove checker names and file names.

[craftmaster1231]

Still filename in commit 772987b

[craftmaster1231]

Also, split this branch into two, please. You can just cherry-pick if your changes are independent.

First two problems should go into one branch and last two in another. No need to describe first two problems that much.
Please, don''t name branches like that. No "fix_warnings" or "warnings12345". You can do master_add_asserts and master_fix_types.