refactor fit_lightcurves and feature output

Change Description

Unifies the feature output format to write using the same function.

Part 1 of #78

[x] My PR includes a link to the issue that I am addressing

Solution Description

Previously, the code would write each line individually, creating the comma separated structure on the fly. Instead of writing per line, we generate all of the features and place them in a pandas.DataFrame and use the pandas to_csv function. This unifies our output method, ensure that all the data is uniform and ensures that we fail gracefully in the even of issues with the features. This also prepares us for saving feature results to MongoDB in a future update.

Code Quality

[x] I have read the Contribution Guide
[x] My code follows the code style of this project
[x] My code builds (or compiles) cleanly without any errors or warnings
[x] My code contains relevant comments and necessary documentation

Project-Specific Pull Request Checklists

Other Change Checklist

[x] Any new or updated docstrings use the NumPy docstring format.
[x] I have updated the tutorial to highlight my new feature (if appropriate)
[x] I have added unit/End-to-End (E2E) test cases to cover any changes
[x] My change includes a breaking change
- [x] My change includes backwards compatibility and deprecation warnings (if possible)

Before [e42321b3]	After [4d964c91]	Ratio	Benchmark (Parameter)
148±2ms	150±2ms	1.02	benchmarks.time_learn_loop('KNN', 'RandomSampling')
140±5ms	141±7ms	1.01	benchmarks.time_feature_creation
197M	197M	1	benchmarks.peakmem_learn_loop('KNN')
188M	188M	1	benchmarks.peakmem_learn_loop('RandomForest')
153±1ms	153±2ms	1	benchmarks.time_learn_loop('KNN', 'UncSampling')
2.56±0.01s	2.56±0.02s	1	benchmarks.time_learn_loop('RandomForest', 'RandomSampling')
2.58±0.02s	2.58±0.01s	1	benchmarks.time_learn_loop('RandomForest', 'UncSampling')

LSSTDESC / RESSPECT