feat: Full chain streaming, increment prototypes

[DaveSkender]

Description

As the next step in our v3 preview release plan, this implements:

• refactor, simplify quote provider (observable only)
• stabilize full prototypes of the core streaming scenarios
  • feat: Quote in » chain out - ADL #1198
  • quote/chain in » chain out - EMA,SMA
  • quote-only in » unchainable result out - ATR-STOP
  • quote/chain in » unchainable result out - ALLIGATOR
  • feat: Quote in » quote/chain out - Renko #1206

Problems to resolve

• get full chain observation to work in isolated EMA/SMA base cases, proof of concept for stream through full chains #1096
• refactor to obviate dual cache, and other EMA/SMA performance tuning
  • refactor performance tests to group Static or Stream, remove redundant series tests from stream grouping
  • abstract QuoteObserver, ChainObserver, and generic Cache methods with generic <T> result storage
• add missing use cases for dispositions Acts, rebuild / cache busting scenarios, in preparation for wider implementation of handle out-of-sequence late-arrival quotes #1094
• allow non-chainable IResult to be an observable » all result forms should be observable, even if they're not chainable (for comprehensive end-use streaming).
• consider: forcing concrete types for observables, cached quotes; I don't think record struct is going to work well with larger types, and has too many stack copies (stack overflow risks). Allowing TQuote is only problematic on observing and caching; maybe just do a quick transform OnArrival();

Additional testing

• test: Stack overflow scenarios, others #1208
  • change forward-cascading updates to use iterative Resend() instead of recursive OnNext()
  • one QuoteProvider with all indicators subscribed (massive parallel)
  • one QuoteProvider with max possible chain depth (massive serial)
• find and fix other, hidden fixed-period initialization indexes

Refactor interfaces

• remove all forms to Tuple interfaces. These are no longer needed due to wider use and compatibility to the IReusable interface; further reducing re-writing transforms and memory utilization.
• convert all public class types to record struct and remove unneeded AbstractResult class. Use of struct type and IObservable has largely limited inheritance options.
  • refactor all uses to "write once", since these are now immutable value types.

Documentation

• update Indicator source taxonomy #1069 for v3, if changed
• update example projects

[DaveSkender]

• add Date alias with [Obsolete], mapped to IQuote.Timestamp
• same for all ISeries classes

[DaveSkender]

Not sure why I keep flip-flopping on the [Serializable] attribute. Microsoft has deprecated all BinaryFormatter related code in .NET 8 due to security flaws.

I think it’s still relevant for other distributed asynchronous messaging frameworks, despite its ill-fated relationship to BinaryFormatter, worth reviewing again to natively support more user environments.

• add [Serializable] attribute to custom exception types
• check to see if this is even needed for new record struct types
• add to any other uncovered return types, if needed

[DaveSkender]

For consideration; but may want to implement this in a v3.1

    /// <summary>
    /// Optional.  Use externally provided cache.
    /// </summary>
    /// <param name="externalCache"></param>
    /// <remarks>
    /// Allow users to provide their own cache storage location.
    /// </remarks>
    protected AbstractCache(
        List<TSeries> externalCache)
    {
        Cache = externalCache;
        throw new NotImplementedException();
    }

[DaveSkender]

• restore MacOS to build matrix, investigate StackOverflowException

[DaveSkender]

• add QuoteProvider() that defaults to built-in Quote class? In addition to QuoteProvider<T>().

• consider using only QuoteProvider() if struct paradigm doesn’t work well with memory management and stack overflow, and we need to revert to record class-only approach, which can’t be well constrained for generic T.

[DaveSkender]

relative speed, on local PC)

Before these changes:

Method	Mean	Error	StdDev
EmaHub	33.77 us	0.644 us	0.689 us
SmaHub	34.77 us	0.482 us	0.451 us

Current state:

Method	Mean	Error	StdDev
EmaHub	1.783 ms	0.0107 ms	0.0089 ms
SmaHub	1.766 ms	0.0156 ms	0.0130 ms

» significant performance refactoring needed
Current changes (needs performance tuning)