Fix bonecache accepting validtimes set in the future.

[zxz41]

Description

As mentioned in https://github.com/ValveSoftware/Source-1-Games/issues/6387, the bone cache will sometimes store a bad timestamp for validity.

This happens (iirc) whenever I.E. C_BaseAnimating::TestHitboxes gets called during interpolation, causing the updated bone cache to set it's valid timestamp to a future tick, which will cause subsequent cache checks (from I.E. the particle system) to assume the cache is still valid, even though it's not.

Performance check

Obviously this will have a slight performance impact because:

• It now needs to do an extra comparison check.
• The more bad future times, the more bone setups because we can't trust the valid time.

I did do some performance checks on a listen server against 24 TFBots on koth_viaduct w/ nb_update_frequency lowered to 0.2 and net_fakelag set to 100 for a single round (roughly 3-6 minutes.)

NOTE this was taken from when the code was doing a similar full check like the server (minus the BONE_USED_BY_BONE_MERGE boneMask.)

-- Profile scopes sorted by time (without children) --
  Scope                               Calls Calls/Frame  Time+Child    Pct        Time    Pct   Avg/Frame    Avg/Call Avg-NoChild        Peak
  ----------------------------- ----------- ----------- ----------- ------ ----------- ------ ----------- ----------- ----------- -----------

-- CURRENT CODE --
    C_BaseAnimating::SetupBones     3173484      35.297   23482.872   8.05   15153.820   5.19       0.261       0.007       0.005       1.109
  C_BaseAnimating::GetBoneCache       85949       0.956     803.309   0.28      59.774   0.02       0.009       0.009       0.001       0.473

-- UPDATED CODE --
    C_BaseAnimating::SetupBones     3130401      45.331   21799.852   8.69   13954.061   5.56       0.316       0.007       0.004       1.130
  C_BaseAnimating::GetBoneCache       82985       1.202     757.432   0.30      57.675   0.02       0.011       0.009       0.001       0.520

Regression check

As far as I can tell, this don't have any regression issues, I've tried it on a bot carrying the intelligence. I've tried it with a bot wearing an unusual via. itemtest. So far I've not been able to notice anything visual and/or game breaking.

Also yes I've read and understood the contributing agreement.