Lower bound improvement based on sharpness of the objective

[dhendryc]

I added the bound improvement using sharpness like outline in the paper.

Written simple test on the Normbox (alias Approx Planted Point) example.

It seems one has to use the node dual gap for the sharpness improvement. I left the code for the root dual gap in for now until we have rechecked the proof in the paper.

[codecov]

Codecov Report

Attention: Patch coverage is 95.65217% with 9 lines in your changes missing coverage. Please review.

Project coverage is 86.87%. Comparing base (ee5cc6c) to head (df13669).
Report is 31 commits behind head on main.

Files with missing lines	Patch %	Lines
src/tightenings.jl	95.63%	9 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #195      +/-   ##
==========================================
+ Coverage   86.23%   86.87%   +0.63%     
==========================================
  Files          17       19       +2     
  Lines        1482     1554      +72     
==========================================
+ Hits         1278     1350      +72     
  Misses        204      204              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dhendryc]

Halved the sharpness constant M analogously to the strong convexity case.