Core implementation for high-performance unique identifiers.

Provides functions for generating and manipulating Flakes - 192-bit identifiers
that combine nanosecond timestamps with high-entropy random data to ensure
both uniqueness and monotonic ordering properties.

High-performance time-ordered unique identifier implementation.

A Flake is a 192-bit identifier combining:
- 64-bit nanosecond-precision timestamp
- 128-bit random component

Designed for distributed systems requiring:
- Monotonic ordering properties
- High collision resistance
- Efficient string serialization

Implementation inspired by μ/log (mulog) by Bruno Bonacci:
https://github.com/BrunoBonacci/mulog/blob/master/mulog-core/src/com/brunobonacci/mulog/flakes.clj
Licensed under Apache License 2.0

High-performance time-ordered unique identifiers.

Flakes are 192-bit identifiers designed for distributed systems that need:
- Guaranteed ordering based on creation time
- High entropy for uniqueness
- Compact string representation
- Fast generation (sub-50ns target)

## Structure
- 64 bits: High-precision timestamp (nanosecond resolution)
- 128 bits: Cryptographically random data

## Key Properties
- **Monotonic**: Later-generated flakes sort after earlier ones
- **Order-preserving**: String representations maintain sort order
- **URL-safe**: Uses web-safe Base64-variant encoding
- **Collision-resistant**: 128 bits of entropy per timestamp

## Performance
Optimized for high-throughput scenarios with thread-local randomization
and efficient encoding algorithms.

High-precision timestamp provider combining wall clock and monotonic timing.

This implementation anchors to the system wall clock at initialization,
then uses monotonic timing to provide nanosecond-precision timestamps
that maintain consistent intervals.

Key characteristics:
- Nanosecond resolution for timestamp generation
- Monotonic properties within process lifetime
- Wall-clock anchored for cross-system compatibility
- Single global instance for consistency

Note: This is not a high-precision wall clock replacement. While it provides
nanosecond granularity, it can drift from true wall time over extended periods
as monotonic clocks don't adjust for NTP synchronization.

Timestamp range extends to approximately year 2262 when using nanosecond
precision in a long value.