Fix display of execution time in datafusion-cli

[Dandandan]

Which issue does this PR close?

Closes #511

Rationale for this change

What changes are included in this PR?

Before this change, starting time was calculated in the printing code, always resulting in 0 seconds.
It also shows the number of seconds with 3 decimals now, which is useful for shorter running queries:

select l_orderkey from X where l_comment like '1%';
0 rows in set. Query took 0.147 seconds.

Are there any user-facing changes?

[codecov-commenter]

Codecov Report

Merging #514 (96b1e34) into master (ac9d4ae) will increase coverage by 0.16%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #514      +/-   ##
==========================================
+ Coverage   75.92%   76.09%   +0.16%     
==========================================
  Files         154      155       +1     
  Lines       26381    26575     +194     
==========================================
+ Hits        20031    20221     +190     
- Misses       6350     6354       +4     

Impacted Files	Coverage Δ
datafusion-cli/src/lib.rs	0.00% <0.00%> (ø)
datafusion-cli/src/main.rs	0.00% <0.00%> (ø)
datafusion/src/optimizer/constant_folding.rs	91.31% <0.00%> (-0.38%)	⬇️
datafusion/src/sql/planner.rs	84.25% <0.00%> (-0.12%)	⬇️
python/src/dataframe.rs	0.00% <0.00%> (ø)
datafusion/src/test_util.rs	96.49% <0.00%> (ø)
datafusion/src/optimizer/utils.rs	48.22% <0.00%> (+0.70%)	⬆️
datafusion/src/sql/utils.rs	66.79% <0.00%> (+1.86%)	⬆️
datafusion/src/physical_optimizer/pruning.rs	92.27% <0.00%> (+2.18%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ac9d4ae...96b1e34. Read the comment docs.

[msathis]

Looks Good

[jimexist]

doing work too fast has its own downside