perf(script): Reduce cost of ScriptEngine::findScript() by around 85%

[Mauller]

This PR dramatically improves script search performance by removing the allocation overhead when comparing script names.

I also tested the same performance degradation against a version of asciistring that does not use the global locks.
But like with the similar waypoint names fix #2761, returning constant references to the asciistrings was significantly more performant.

All of these tests were performed in a retail release, letting the same AoD map level run to its conclusion.
So this is not a debug only issue.

---

Flame graphs and function usage comparisons.

Before:

After:

---

video comparison.

This video shows the most recent GO build against a tweaked GO build with the fixes from this PR.

https://youtu.be/aDy8w4sHPKU

---

TODO

• Replicate in generals

[Caball009]

const got a bit out of hand; use these:

const AsciiString& getName() const { return m_groupName;}
const AsciiString& getName() const { return m_scriptName;}
const AsciiString& getConditionTeamName() const {return m_conditionTeamName;}

---

Did removing the global lock from the AsciiString member functions provide less performance improvement than these changes? That'd be surprising to me.

[bobtista]

I found one issue a few times, otherwise this looks good to me, and wow on the performance boost - great find :)

btw potential follow-up after this merges: findScript/findGroup are still O(total scripts) linear scans across every side, called per lookup.

Edit: Just noticed Caball had already found these - sorry for duplicating!

[Mauller]

const got a bit out of hand; use these:

const AsciiString& getName() const { return m_groupName;} const AsciiString& getName() const { return m_scriptName;} const AsciiString& getConditionTeamName() const {return m_conditionTeamName;}

Did removing the global lock from the AsciiString member functions provide less performance improvement than these changes? That'd be surprising to me.

Yeah, removing the global lock did provide less performance compared to passing references. The same happened with the waypoint and trigger labels fix.

[Mauller]

Tweaked and ready to go, i have no idea why some checks are not running though.

[Caball009]

It probably detects you didn't make changes to the Generals code yet.

[Mauller]

It probably detects you didn't make changes to the Generals code yet.

I did not realise that change had been implemented.

[xezon]

Looking good