Create new module for handling doc int responses to replace bloated `Utility.ts`

What/Why

moving the DocIntResponse interaction into its own module creates reusable code
the existing file is poorly organized, badly named, and in need of refactor
this will also bring us two major functionality upgrades:
- per-word visual bounds
- user selection that does not rely on the text layer and thus is not subject to loss of fidelity

TODO

[x] rangeToSummary
- [ ] tests
[x] excerptToSummary - already written and needs refactor
- [ ] tests
[ ] setDocumentDelta
- [ ] once this function exists, plumb delta through remaining functionality
- [ ] tests
[x] createRegions
- [ ] tests
[ ] conversion between polygonC and polygon..N? for draw function usage

Types & Interfaces

docint standard with the following modifications:

interface AnalyzeResult {
  apiVersion: string;
  modelId: string;
  stringIndexType: string;
  content: string;
  delta?: number; // <- added by Preprocess.ts
  pages: Page[];
  tables: Table[];
  paragraphs: Paragraph[];
  styles: Style[];
  contentFormat: string;
  sections: Section[];
  figures: Figure[];
}

export interface Page {
  pageNumber: number;
  angle: number;
  width: number;
  height: number;
  unit: string;
  words: Word[];
  lines: Line[];
  regions?: Region[]; // <- added by Preprocess.ts, see below for definition
  spans: Span[];
  selectionMarks?: SelectionMark[];
}

Additionally, all cases of polygon: number[] have been changed to polygon: polygon4 to use the new type below:

export type polygon4 = [
  number, number,
  number, number,
  number, number,
  number, number
];

// This will be the output of a conversion function to convert polygonC into pure number[] polys
export type polygonN = number[];

// This does mean supporting cases where the lead and
// the tail are non-adjacent (i.e. the selection travels
// to the next line but not far enough to overlap with
// the previous line)
// might need some basic typescript help here to express what I want happening:
// lead always exists, then tail might exist, *then* body might exist (iff tail exists)
export type polygonC =
  | {
      lead: polygon4;
      tail?: polygon4;
    }
  | {
      lead: polygon4;
      body?: polygon4;
      tail: polygon4;
    };

// lowercase polygon is pageless
export type polygon = polygon4 | polygonC

polygon4 is explicitly a 4-pt poly, while polygonC is a complex poly. Still deciding whether Summaries (see below) always explicitly return polygonC or if they can return either.

More interfaces:

// uppercase Polygon has a page
export interface Polygon {
  polygon: polygon;
  page: number;
}

// lowercase point is pageless
export interface point {
  x: number;
  y: number;
}

// uppercase Point has a page
export interface Point {
  point: point;
  page: number;
}

// lowercase range is just number primitives and is pageless
// represents any start, end range
export type range = [number, number]

// uppercase range uses Points and therefore has pages
// specifically represents a CURSOR range not a drawn box
export interface Range {
  start: Point;
  end: Point;
}

// Akin to columns or sections
// not relevant to the Viewer etc, just used in-module
export interface Region {
  polygon: polygon4;
  lineIndices: range;
  wordIndices: range;
}

// an excerpt and its location(s)
export interface Summary {
  excerpt: string;
  polygons: Polygon[];
}

Functions

Preprocess.ts

// in place change to di, adds a field to di.analyzeResult that models the spacing of the document (scale of 0-1?)
export function setDocumentDelta(
  di: DocumentIntelligenceResponse
);

// in place change to di, adds a `Region[]` field to each page of analyzeResult
export function createRegions(
  di: DocumentIntelligenceResponse
);

DocInt.ts (naming?)

// takes in a cursor range (aka, assumed to be contiguous in the document; we're not drawing a box)
// and creates the corresponding Summary
export function rangeToSummary(
  range: Range,
  di: DocumentIntelligenceResponse
): Summary;

// takes in an excerpt and creates the corresponding Summary with the *first* instance of the excerpt in di
export function excerptToSummary(
  excerpt: string,
  di: DocumentIntelligenceResponse
): Summary;

billba / excitation