Improvement: Add slice-based calculation methods for high-performance batch processing

[cinar]

Description

Currently, all indicators in the library are tightly coupled to channels and stream-based processing. While channels are excellent for real-time streaming data, batch operations (such as running backtests over historical datasets loaded from CSVs) can incur scheduling, allocation, and context-switching overhead from spawning multiple goroutines and passing values through channels.

To improve performance for batch calculations and backtesting, we should introduce a way to compute indicators directly on slices with zero goroutines.

Proposed Changes

1. Define a stateful calculation interface (e.g., Calculator or similar step-by-step interface) representing the state of an indicator:

   type Calculator[T Number] interface {
       Next(input T) (output T, ok bool)
   }

2. Implement this interface on existing indicator structs (like Sma, Ema, etc.).
3. Add a new ComputeSlice(inputs []T) []T method to the indicators, which processes the inputs sequentially in a simple loop without channels.
4. Keep the existing channel-based Compute(<-chan T) <-chan T methods intact for backwards compatibility, but rewrite their internals to utilize the stateful calculator logic.