[Draft] Investigating SQL Support. (#11)

Beginning to investigate SQL support. Right now, for anything beyond a
small Xarray dataset, this is terribly slow.
This PR also introduces a profiling methodology using py-spy
(statistical profiling).

perf_tests/README.md

@@ -0,0 +1,19 @@
+# Performance testing & profiling
+
+So far, this includes statistical profiles via py-spy. 
+
+## Dev Process
+
+1. Run a profile test with the `profile.sh` script as so:
+
+   ```shell
+   # PROFILE_CASE_PY=groupby_air.py
+   sudo ./profile.sh $PROFILE_CASE_PY
+   ```
+
+   This will open a flame graph in the browser.
+
+2. After tuning code in qarray, run another profile to generate a SVG.
+
+3. Please commit the "after" profile SVG along with the performance improvements.
+

perf_tests/groupby_air.py

@@ -0,0 +1,34 @@
+#!/usr/bin/env python3
+
+import xarray as xr
+import qarray as qr
+from dask_sql import Context
+
+
+if __name__ == '__main__':
+  air = xr.tutorial.open_dataset('air_temperature')
+  chunks = {'time': 240, 'lat': 5, 'lon': 7}
+  air = air.chunk(chunks)
+  air_small = air.isel(
+    time=slice(0, 12), lat=slice(0, 11), lon=slice(0, 10)
+  ).chunk(chunks)
+
+  df = qr.to_dd(air_small)
+
+  c = Context()
+  c.create_table('air', df)
+
+  query = c.sql('''
+      SELECT
+        "lat", "lon", SUM("air") as air_total
+      FROM 
+        "air" 
+      GROUP BY
+       "lat", "lon"
+      ''')
+
+  result = query.compute()
+
+  expected = air.dims['lat'] * air.dims['lon']
+  assert len(result) == expected, f'Length must be {expected}.'
+  print(expected)