new rolling functions: min and prod

[jangorecki]

Two new rolling functions, towards #2778.
Supersede #5682

Optimizations

• for min are the same as used for max
• for prod are the same as used for mean

---

benchmark vs very fast roll package

library(roll)
library(data.table)
set.seed(108)
x = rnorm(1e8)
n = 1e3

## PROD
system.time(a <- roll_prod(x, n))
#   user  system elapsed 
#  1.083   0.836   1.919
system.time(b <- frollprod(x, n))
#   user  system elapsed 
#  0.822   0.384   1.209 
all.equal(a,b)
#[1] TRUE

## MIN
system.time(a <- roll_min(x, n))
#   user  system elapsed 
#  1.896   0.840   2.739
system.time(b <- frollmin(x, n))
#   user  system elapsed 
#  0.278   0.394   0.672
all.equal(a,b)
#[1] TRUE

This is for rnorm, be aware that for specific input, data.table's implementation can be much slower, see issue reported for rolling max #5923

---

This work was supported by the Medical Research Council, UK [grant number 1573].

[github-actions]

• HEAD=froll-min-prod stopped early for isoweek improved in #7144
  

Generated via commit 227b39e

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	2 minutes and 59 seconds
Installing different package versions	21 seconds
Running and plotting the test cases	2 minutes and 38 seconds

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.09%. Comparing base (ed2f36f) to head (227b39e).
⚠️ Report is 4 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #7299      +/-   ##
==========================================
+ Coverage   99.08%   99.09%   +0.01%     
==========================================
  Files          83       83              
  Lines       15781    16052     +271     
==========================================
+ Hits        15636    15907     +271     
  Misses        145      145              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ben-schwen]

not a big fan of macros but couldn't we do something like and deduplicate wmin and wmax?

#define DEFINE_WINDOWFUN(fname, init_val, op)
DEFINE_WINDOWFUN(wmax, R_NegInf, >=)
DEFINE_WINDOWFUN(wmin, R_PosInf, <=)

[jangorecki]

could be, one reason not to is reduced code coverage

[ben-schwen]

Same holds for frollmaxFast and frollmaxExact

[jangorecki]

Same holds for frollmaxFast and frollmaxExact

frollmax Fast vs. Exact does not have that much in common, you possibly meant froll Min vs Max

as for the main functions, to turn them into macros would require __func__ handling as well. Additionally it would reduce clarify of code, and friendliness for code editors.

As for the helper functions wmin, wmax, IMO it indeed make sense, but maybe better as a follow up. So this PR will be focused more on adding new things rather than refactoring previous ones.