Add Demo-2

[abhi3700]

Description

This PR adds code + data for this logic:

Consider sample dataset of 1000 reviews.

1. For each sample in dataset (use the 1000 reviews)
   1. Create embedding of sample
   2. Save embeddings to disk, linked with source sample
2. For each embedding in embeddings
   1. for each nbits = [8, 16, 32, 64, 128]
      1. Apply LSH to the embedding
      2. Add hash to a hash table, along with index of embedding
      3. Save hashes to disk (categorize by parameter)

Here is the output file (in CSV format) with following columns:

Text	Embedding	Hash 8-bit	Hash 16-bit	Hash 32-bit	Hash 64-bit	Hash 128-bit

[abhi3700]

One observation when repeating data pre-processing, for most of the text reviews out of 1000, the values get changed after 3-4 decimal places. 63d48d9

I don't think it would affect in bucketing.

[abhi3700]

This is the error at when calling function: pl.read_csv() directly…

polars.exceptions.ComputeError: could not parse `00011011110011110011001001110010` as dtype `i64` at column 'Hash 32-bit' (column number 5)

The current offset in the file is 34598 bytes.

You might want to try:
- increasing `infer_schema_length` (e.g. `infer_schema_length=10000`),
- specifying correct dtype with the `dtypes` argument
- setting `ignore_errors` to `True`,
- adding `00011011110011110011001001110010` to the `null_values` list.

Original error: ```remaining bytes non-empty```

Although this can be fixed by reading as str instead of default i64 type. Not sure if there is any solution in vector DB or not 🤔 .
Now, it’s fixed. commit

[abhi3700]

Observation (for 20 paragraphs): The matrix does have the relatively lowest Hamming distance diagonally as expected. You can find those in matrix_8.csv, matrix_16.csv, etc.

[abhi3700]

Added interactive (in html format…..hover on it for values) heatmap plots to give a visualization of the generated matrices for each nbits = 8, 16, 32, 64, 128 for 20 paragraphs b/w original & slightly modified texts. Download the html plots from commit.
👀 With increasing no. of hyperplanes, the diagonal blue line is more crystal clear.