Create correct xlim for ts plot and update datetimelike tests

[charlesdong1991]

xref: #27993

The view limit for non-datetime plotting is solved. and behavior will be the same as 0.23.4
e.g.

HOWEVER, issue in datetime-like still exists:
The issue with datetime-like plotting, xlim will be changed, and the view limit is set to data limit, so there is no margin on xaxis. Plots are looking like:

[TomAugspurger]

@charlesdong1991 I copied your comment from #27993 (comment) into the original issue. Hope that's OK.

Do you still have the actual code snippets, rather than the text? It'd be nice to have them around.

I do agree that the "after your change" behavior looks correct.

[charlesdong1991]

ahh, thanks for editing, I should have done more on description. @TomAugspurger

what do you mean by actual code snippets? something like this? (this was basically copied from the actual tests from test_datetimelike.py)

yrs = [3.5, 11]

xpl1 = xpl2 = [pd.Period("1988Q1").ordinal] * len(yrs)
rs1 = []
rs2 = []
for i, n in enumerate(yrs):
    rng = pd.period_range("1987Q2", periods=int(n * 4), freq="Q")
    ser = pd.Series(np.random.randn(len(rng)), rng)
    _, ax = plt.subplots()
    ser.plot(ax=ax)
    xaxis = ax.get_xaxis()
    rs1.append(xaxis.get_majorticklocs()[0])

    (vmin, vmax) = ax.get_xlim()
    ax.set_xlim(vmin + 0.9, vmax)
    rs2.append(xaxis.get_majorticklocs()[0])

[charlesdong1991]

and PR for this test is coming!

[charlesdong1991]

i am sincerely apologizing for this, I realize today when doing tests that my fix is improper/wrong in terms of timeseries plotting. I am still trying to find root cause, hope to correct it asap before new release. Sorry again for it. @TomAugspurger
the issue/problem is when plotting datetime like, it will lose the pandas datetime formatter information on xaxis. if it's not datetime, then it will be fine.

[arthurire]

Just want to make sure if you have fixed this issue in 0.25.2, I can still reproduce this error in 0.25.2.

[charlesdong1991]

Hi, thanks for your comment, @arthurire However, I could not reproduce it in master, this looks fine to me.

[charlesdong1991]

emm, unfortunately i have to take back my previous comment.

Just found out that this works as expected in my development setup, but if I import pandas in other environments/places, then I could reproduce this issue. Looks weird.