Implement Data Sources

[nagakingg]

Description

Addresses #287 -- Implements DataSource, SimAsset, TimeSequence

New classes:

• Add DataSource, SimAsset, and TimeSequence template classes
• Add SimAssets: OnChainAsset, OnChainAssetPair
• Add TimeSequence: DateTimeSequence
• Add DataSources: CoinGeckoPriceVolumeSource, CsvDataSource

Changes to incorporate new classes:

• Added get_asset_data() and get_pool_data() convenience functions to pipelines.common
• Added pool_data.get_pool_assets()
• Modified interfaces to pipeline functions, get_price_data(), get_pool_volume() for consistency with new class interfaces
• Moved price/volume data retrieval outside of PriceVolume iterator

Removed because no longer needed:

• Removed SimAssets type and SimPool.assets property
• Removed coin_names property from PricingMetrics
• Removed Coingecko pool_prices and coin_ids_from addresses

See changelog for more details.

Fixes to CoinGecko Data (!):

• Changed Coingecko price data resampling to hourly samples with 10 minute tolerance. Previously, we sampled at the 30-minute mark between hours to capture late timestamps (e.g. resample to 1:30, 2:30, 3:30 to capture data timestamped at 1:02, 2:17, & 3:06). Coingecko's data timing has become more reliable, so a 10-minute tolerance locked to the hour is more accurate. Note that this breaks the end-to-end tests for this PR, since the resampling of the data is altered
• Changed end-to-end tests to use more recent data with more reliable timestamps
• Fixed a bug where volume units were converted incorrectly for Coingecko data. Volumes were assumed to be in units of the queried asset, but I verified that they were actually in dollars. This bug was introduced by me in commit df79810. This has not affected any Curve optimizations since we have always used Kaiko or other custom data since then, but it may have affected some users.

Hygiene checklist

• Changelog entry
• Everything public has a Numpy-style docstring
  (modules, public functions, classes, and public methods)
• Commit history is cleaned-up with minor changes squashed together
  and descriptive commit messages following Tim Pope's style

Cute Animal Picture

[chanhosuh]

Overall things are looking very nice. I think there could be more things done on the csv data source, but those can be saved for future PRs.

[nagakingg]

Here is a results comparison using the simple pipeline. Current main (_old.html) vs this PR (_new.html).

Everything looks nearly the same except for triCRV annualized returns (in crvUSD). Examining the pool value time-series, this is just due to the old version including an initial timestamp with significantly lower prices. In other words, this looks like just an artifact of the precise sample timing being slightly different.

The newer timestamp generation (using TimeSequence) is a bit more refined than the old one, so I see no need to try to replicate the old test results explicitly.

results_comparison.zip

Naga has addressed the feedback